Cosmo support #2013

mkeeter · 2025-02-24T20:14:06Z

I'm opening this early to run it through CI and do preliminary self-review, but don't feel obliged to look at it yet

nathanaelhuffman · 2025-03-24T16:15:28Z

drv/cosmo-seq-server/src/main.rs

+                if !okay {
+                    // We'll return to A2, leaving jefe and our local state
+                    // unchanged (since they're set after this block).
+                    self.log_pg_registers();


It would be nice to know what seq API state and seq raw state we were in at the point of timeout, in addition to the PGs. Combining these pieces of information can help us pinpoint which rail(s) didn't come up or let us figure out what we are waiting on.

Sure, I added a call to self.log_state_registers() here as well (and below) in 128d6f9

drv/cosmo-seq-server/src/main.rs

bcantrill

Looking good; two potentially incorrect values plus some questions/nits.

app/cosmo/base.toml

task/power/src/bsp/cosmo_a.rs

task/thermal/src/bsp/cosmo_a.rs

mkeeter · 2025-03-27T13:55:39Z

task/thermal/src/bsp/cosmo_a.rs

+
+        // Bonus bits for M.2 power, which is switched separately.  We *cannot*
+        // read the M.2 drives when they are unpowered; otherwise, we risk
+        // locking up the I2C bus (see hardware-gimlet#1804 for the gory


We should check and see if this is still necessary!

It should be unnecessary, so I removed it (see oxidecomputer/quartz#321 and https://github.com/oxidecomputer/hardware-cosmo/issues/641)

mkeeter · 2025-03-28T13:14:12Z

task/thermal/src/bsp/cosmo_a.rs

+    sensors::NUM_NVME_BMC_TEMPERATURE_SENSORS;
+
+// The control loop is driven by CPU, NIC, and BMC temperatures
+// XXX we should also monitor DIMM temperatures here


@nathanaelhuffman / @Aaron-Hartwig I think we need integration with the main FGPA here, because we'll be reading DIMM temperatures through the I3C proxy.

This is true though may be a post-bringup activity. We've been having a number of issues with the ruby+grapefruit signal integrity to properly prototype this and while I'm still working to improve the test fixture over the next couple of days, we're running out of time before lab pack-up must commence.

labbott · 2025-04-29T14:17:59Z

app/cosmo/base.toml

+task-slots = ["sys"]
+notifications = ["spi-irq"]
+
+# XXX this is only used by cosmo_seq; could we merge it?


XXX comment

This remains a potential future improvement, but I don't want to handle it right now!

app/cosmo/src/main.rs

labbott · 2025-04-29T14:21:07Z

build/fpga-regmap/src/lib.rs

+                    1 => unreachable!(),
+                    2..=8 => "u8",
+                    9..=16 => "u16",
+                    17..=32 => "u32",
+                    _ => panic!("invalid width {width}"),


it feels a little strange to have 1 be unreachable but everything else panic

They are meaningfully different, though! We should never hit the 1-bit path, because cases where lsb == msb are handled as bool. I'll add some text to the unreachable!(..) to make that more clear.

labbott · 2025-04-29T14:24:03Z

drv/cosmo-seq-server/src/main.rs

+
+        use fmc_periph::A0Sm;
+        match (self.get_state_impl(), state) {
+            (PowerState::A2, PowerState::A0) => {


Did we want to add the CPU_PRESENT CORETYPE SP5 checks before merging? e.g. https://github.com/oxidecomputer/hubris/blob/master/drv/gimlet-seq-server/src/main.rs#L763

Opened #2051 in the interest of getting this merged

drv/spartan7-loader/src/main.rs

task/packrat/src/cosmo.rs

labbott · 2025-04-29T14:28:04Z

task/power/src/bsp/cosmo_a.rs

+/// TODO: explain rationale for this value.
+const TRACE_DEPTH: usize = 52;


Copy-pasta from Gimlet 🙃

labbott · 2025-04-29T14:28:31Z

task/thermal/src/bsp/cosmo_a.rs

+                gain_i: 0.0135,
+                gain_d: 0.4,
+                min_output: 0.0,
+                max_output: 10.0, // XXX fix this before merging


Fix this before merging

@nathanaelhuffman now that we're getting CPU temperatures, can we bump this back up to 100%?

labbott · 2025-04-29T14:28:45Z

task/thermal/src/bsp/cosmo_a.rs

+}
+
+// In general, see RFD 276 Detailed Thermal Loop Design for references.
+// TODO: temperature_slew_deg_per_sec is made up.


Is this still true?

Absolutely!

labbott

⛵

mkeeter force-pushed the mkeeter/cosmo-support branch 4 times, most recently from 22a84bf to 7c30d67 Compare February 25, 2025 14:59

mkeeter changed the base branch from master to mkeeter/countable-ice40-errors February 25, 2025 14:59

Base automatically changed from mkeeter/countable-ice40-errors to master February 25, 2025 16:18

mkeeter force-pushed the mkeeter/cosmo-support branch from dfbb462 to ab18423 Compare February 25, 2025 21:27

mkeeter force-pushed the mkeeter/cosmo-support branch 3 times, most recently from 73fbbb9 to e9d1098 Compare March 11, 2025 13:50

mkeeter force-pushed the mkeeter/cosmo-support branch from e9d1098 to a25bac1 Compare March 12, 2025 14:14

mkeeter mentioned this pull request Mar 18, 2025

want Cosmo image #1976

Open

mkeeter force-pushed the mkeeter/cosmo-support branch 2 times, most recently from 2189162 to f8387a6 Compare March 21, 2025 14:23

nathanaelhuffman reviewed Mar 24, 2025

View reviewed changes

mkeeter force-pushed the mkeeter/cosmo-support branch from fff65cd to 8190dfe Compare March 24, 2025 20:13

mkeeter marked this pull request as ready for review March 24, 2025 21:16

bcantrill reviewed Mar 27, 2025

View reviewed changes

mkeeter commented Mar 27, 2025

View reviewed changes

mkeeter commented Mar 28, 2025

View reviewed changes

mkeeter force-pushed the mkeeter/cosmo-support branch from 9c367eb to 4a13b4e Compare April 14, 2025 19:11

nathanaelhuffman force-pushed the mkeeter/cosmo-support branch from 608fcab to e91e4f9 Compare April 23, 2025 02:46

mkeeter mentioned this pull request Apr 23, 2025

Scaling on MAX5970 Current Sense is incorrect. Need some register settings changed as well #2047

Closed

nathanaelhuffman force-pushed the mkeeter/cosmo-support branch from c57a592 to 2c3bde0 Compare April 23, 2025 23:17

labbott reviewed Apr 29, 2025

View reviewed changes

mkeeter and others added 5 commits April 29, 2025 10:50

Cosmo implementation work

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.

GPG key ID: 4AEE18F83AFDEB23
Expired

Verified
Learn about vigilant mode

048ad49

Add host NMI support

49e1028

Add fault and chassis LED handling

419b746

Adding cosmo-seq inital files and structure

ad8e96c

Update cosmo-hp and move into subfolder

3ae1805

nathanaelhuffman and others added 25 commits April 29, 2025 10:50

Really sequences, with the real fix

aee0f82

Switch host flash back to SP when going -> A2

af8427c

Working boot into OS

c729166

I need assistance

e24aa6b

Let FMC demo poke everywhere in MMIO space

4eba02b

Working hotplug

fbb9578

Clamp PWM output

c78c951

Use pid_config.max_output instead of hard-coded 100 in full-power states

e354c0e

Make SP_TO_HOST_CPU_INT_L push-pull on Cosmo only

b39df4f

Make SP_TO_SP5_NMI_SYNC_FLOOD_L push-pull

bf18bc6

working IPCC, reboots still not working

e3c63b0

Rebooting works, IPCC works, hotplug works

0c60896

Adjust MAX5970 resistance for cosmo sharkfins

5aeca3c

Adjust MAX5970 current threshold for 12V U.2s

6ad2595

Cargo format

dbb2ca7

Add updated FPGAs for hotplug investigation

e32fb5f

Working hotplug

e4c9c0b

tagged actual fpga releases

4744240

Maybe T6 fix?

6c1b0df

Fix edge condition in flash writing

d1f6a2a

Get VPD identity in cosmo sequencer

9f0cdc8

Configure SP_TO_SP5_PROCHOT

d53176a

Fix SBRMI / SBTSI addresses

2205b75

misc FPGA fixes for APML, T6 logic etc and a real quartz release

3f866bf

Fixes from Laura's review

Loading
Loading status checks…

a07a13a

mkeeter force-pushed the mkeeter/cosmo-support branch from 6c6ea7c to a07a13a Compare April 29, 2025 14:50

Bump fan speed

Loading
Loading status checks…

b7101cd

labbott approved these changes Apr 29, 2025

View reviewed changes

mkeeter merged commit cc3ff37 into master Apr 29, 2025
135 checks passed

mkeeter deleted the mkeeter/cosmo-support branch April 29, 2025 15:06

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Cosmo support #2013

Cosmo support #2013

mkeeter commented Feb 24, 2025

nathanaelhuffman Mar 24, 2025

mkeeter Mar 24, 2025

bcantrill left a comment

mkeeter Mar 27, 2025

mkeeter Mar 28, 2025

mkeeter Mar 28, 2025

nathanaelhuffman Mar 28, 2025

labbott Apr 29, 2025

mkeeter Apr 29, 2025

labbott Apr 29, 2025

mkeeter Apr 29, 2025

labbott Apr 29, 2025

mkeeter Apr 29, 2025

labbott Apr 29, 2025

mkeeter Apr 29, 2025

labbott Apr 29, 2025

mkeeter Apr 29, 2025

labbott Apr 29, 2025

mkeeter Apr 29, 2025

labbott left a comment

		/// TODO: explain rationale for this value.
		const TRACE_DEPTH: usize = 52;

Cosmo support #2013

Cosmo support #2013

Conversation

mkeeter commented Feb 24, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bcantrill left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

labbott left a comment

Choose a reason for hiding this comment