Satellite system doesn't get uplink clock.

[nkrackow]

Bug Report

One-Line Summary

Building a bitstream for a satellite with an ARTIQ version after 47581e0 leads to the satellite not getting a link to the master.

Issue Details

I'm pretty sure 0775ae1, 8f3d06a are the culprit.
If I find time tomorrow I'll check this.
The info output about the io expander is a fix for an unrelated hardware change that is not in upstreamed in ARTIQ yet.
I verified that it has nothing to do with this problem by testing a verbatim ARTIQ earlier but I don't have the console out of that right now.

Expected Behavior

[     0.000003s]  INFO(satman): ARTIQ satellite manager starting...
[     0.005908s]  INFO(satman): software ident 7.0.beta;stockholmsatellite
[     0.012445s]  INFO(satman): gateware ident 7.0.beta;stockholmsatellite
[     0.029525s]  INFO(board_misoc::io_expander): MCP23017 io expander 0 not found. Checking for PCA9539.
[     0.058069s]  INFO(board_misoc::io_expander): MCP23017 io expander 1 not found. Checking for PCA9539.
[     0.456029s]  INFO(board_artiq::si5324): waiting for Si5324 lock...
[     2.408998s]  INFO(board_artiq::si5324):   ...locked
[     2.797873s]  INFO(satman): uplink is up, switching to recovered clock
[     2.830818s]  INFO(board_artiq::si5324): waiting for Si5324 lock...
[     4.559314s]  INFO(board_artiq::si5324):   ...locked
[     7.885235s]  INFO(board_artiq::si5324::siphaser): calibration successful, lead: 198, width: 433 (347deg)
[     8.235495s]  INFO(satman): TSC loaded from uplink
[     8.307408s]  INFO(satman): rank: 1
[     8.309507s]  INFO(satman): routing table: RoutingTable { 0: 0; 1: 1 0; 2: 2 0; 3: 3 0; }

Actual (undesired) Behavior

[     0.000003s]  INFO(satman): ARTIQ satellite manager starting...
[     0.005908s]  INFO(satman): software ident 7.0.beta;stockholmsatellite
[     0.012445s]  INFO(satman): gateware ident 7.0.beta;stockholmsatellite
[     0.029525s]  INFO(board_misoc::io_expander): MCP23017 io expander 0 not found. Checking for PCA9539.
[     0.058069s]  INFO(board_misoc::io_expander): MCP23017 io expander 1 not found. Checking for PCA9539.
[     0.456029s]  INFO(board_artiq::si5324): waiting for Si5324 lock...
[     2.376948s]  INFO(board_artiq::si5324):   ...locked

It never gets an uplink connection after this.

Setup

No flash config file on the satellite.

[nkrackow]

8f3d06a#diff-0c99ae49963f6b70bf1690eaba102795097c1089358b20dcc83ee82422cca57cR120-R121
I don't fully understand this comment but maybe that has something to do with it? @dnadlinger

[dnadlinger]

Your guess isn't unreasonable, but I'm also not sure how the commits would be related – AFAIK they should only affect the DRTIO master.

[dnadlinger]

What are the master's clock settings (that this satellite is supposed to connect to)?

[dnadlinger]

To be clear, the changes only (are supposed to) affect Kasli v2 when used as a master, making it such that internal/external clocks can actually be selected at runtime through the rtio_clock config option without also having to recompile the firmware.

[jordens]

How do you come to the impression that the changes are limited to standalone/master? Maybe this was tested.
0775ae1#diff-3ed5c793f81f4728447227870960117251dafbb1869a0ff428661c7d97fd5384R221
in the context of

artiq/artiq/firmware/satman/main.rs

Line 452 in 874d298

crystal_ref: true

and

artiq/artiq/firmware/satman/main.rs

Line 521 in 874d298

si5324::setup(&SI5324_SETTINGS, si5324::Input::Ckin1).expect("cannot initialize Si5324");

looks like an impact to me.

[dnadlinger]

Yep, that was just a failure of Find All Uses (and apparently, also my subsequent rging) leading me to assume that the siphaser/… code used a different code path for initialisation.

[dnadlinger]

Reopened this until somebody can confirm the issue fixed, as I can't test this locally until next week, and there doesn't seem to be CI coverage for DRTIO satellites.

[sbourdeauducq]

Also need to be cherry-picked on release-7.

[nkrackow]

I've now tested 1dd0d34 on a Kasli V2.0.2 satellite and master and all seems to work as before on both sides.

[dnadlinger]

Thanks!