[HN Gopher] Clusterboard A64 Insidious Reset Problem: Solved
       ___________________________________________________________________
        
       Clusterboard A64 Insidious Reset Problem: Solved
        
       Author : nreilly
       Score  : 76 points
       Date   : 2021-08-01 08:03 UTC (14 hours ago)
        
 (HTM) web link (ericdraken.com)
 (TXT) w3m dump (ericdraken.com)
        
       | tyingq wrote:
       | Wow, that was really involved. The breakthrough seemed to be him
       | seeing the through-hole via with the microscope where no branch
       | in the circuit was supposed to exist. That is, after a whole
       | bunch of dead-end exploratory exercises. Feels like the Pine
       | folks should ship him a thank you package of goodies.
        
       | pm215 wrote:
       | Random nitpick from the middle of a good writeup of tracking down
       | a hardware bug, but 'SMC' in this context doesn't stand for
       | 'Secure Memory Controller'. It's the Arm instruction that lets
       | normal-world (kernel) code call into secure world code
       | (firmware). Just like the SVC insn lets userspace call the
       | kernel, SMC lets the kernel call the trusted firmware. PSCI is a
       | standardized API that the firmware implements with functions like
       | 'power CPU on/off' and 'reset system'. The PSCI spec is public so
       | you can get the function ID values and argument values from there
       | rather than guessing them from source code. PSCI is the preferred
       | mechanism for reset, cpu hotplug, etc on 64-bit Arm (it abstracts
       | the details of prodding reset controller hardware and so on away
       | and lets the firmware deal with it.)
        
         | floatboth wrote:
         | Heh, this was even in the u-boot `smc` command help text
         | snippet in the article.
        
       | justinclift wrote:
       | This is really good news!
       | 
       | I'd considered getting a Clusterboard previously, but the reset-
       | not-actually-working problem was a showstopper.
       | 
       | If this "fix" makes it into the next hardware revision of the
       | Clusterboard, that should mean it's a feasible
       | learning/experimentation platform after all. :)
        
       | robert_foss wrote:
       | _work trauma intensifies_
        
       ___________________________________________________________________
       (page generated 2021-08-01 23:01 UTC)