MTL platforms fail to boot with some memory modules

[mkopec]

Component

Dasharo firmware

Device

NovaCustom V54 14th Gen, NovaCustom V56 14th Gen

Dasharo version

v0.9.0-rc3

Dasharo Tools Suite version

No response

Brief summary

Some memory modules don't work well with MTL laptops.

How reproducible

No response

How to reproduce

with 16GB module:

• booting once after flashing works
• rebooting works
• power off and power back on doesn't work

with 32GB module:

• booting with module in slot 2 works
• booting with module in slot 1 doesn't work

Expected behavior

All modules must work.

Actual behavior

Not all modules work.

Screenshots

No response

Additional context

No response

Solutions you've tried

No response

[marmarek]

with 16GB module:

* booting once after flashing works

* rebooting works

* power off and power back on **doesn't work**

I have similar experience with 48GB module in slot 2, but in my case even rebooting doesn't work.

The same module in slot 1 doesn't work even once (but this I tried on a slightly older build, I can re-test with a recent build if needed).

All tests on V560TU.

[mkopec]

Apparently fixed by using FSP v4122_14. Need to test 48GB modules still.

[wessel-novacustom]

On V540TU with Ultra 5, the laptop doesn't boot after external flashing v0.9.0-rc4 and the belonging EC, then going to the BIOS settings immediately and select Reset (as in: restart).

Powering off and on doesn't bring the device back to life either, the screen display remains off.

So it did boot the first time.

The unit has 1x 16 GB internal memory installed in slot 2, and a PX600 SSD installed in the slot on top of the mainboard.

[marmarek]

Same for me with rc4 - still broken the same way.

[marmarek]

Is there some extra debugging info I can extract to help? Or maybe test some specific procedure?

[wessel-novacustom]

On V560TU with Ultra 7, with v0.9.0-rc4 for this laptop, with 2x 48 = 96 GB internal memory and 2x PX700 4 TB SSD, booting, rebooting and cold boot works fine. Boot time after first memory initialisation is about 8 seconds.

[wessel-novacustom]

On another V540TU with ultra 7 and v0.9.0-rc4 BIOS and belonging EC, with 1x 32 GB internal memory installed and 1x PX700 SSD on the MB, 1x PX600 SSD installed at the battery, the start/boot/restart behaviours are as expected, and the boot time is way better than on the V560TU from the message above.

[mkopec]

Okay, I've reproduced on a V560TU unit. Could not reproduce on a V540TU unit. I get an FSP-S error (Not FSP-M weirdly, so maybe not related to memory after all)

[16:22:00.638] [INFO ]  FSPS returned 80000003
[16:22:00.654] [EMERG]  FspSiliconInit returned with error 0x80000003

(plus some FSP debug logs that I can't paste here)

[wessel-novacustom]

On another V540 with Ultra 5 and v0.9.0-rc4 BIOS and belonging EC, with 1x 32 GB internal memory installed in slot 2, and 1x PX600 SSD on the MB, then going to the BIOS settings immediately and select Reset (as in: restart). Powering off and on doesn't bring the device back to life either, the screen display remains off.

So it did boot the first time. It looks like the chips on the internal memory modules are sometimes different:

I'll now try with the other module to see if the behaviour is the same.

[wessel-novacustom]

On another V540 with Ultra 5 and v0.9.0-rc4 BIOS and belonging EC, with 1x 32 GB internal memory installed in slot 2, and 1x PX600 SSD on the MB, then going to the BIOS settings immediately and select Reset (as in: restart). Powering off and on doesn't bring the device back to life either, the screen display remains off.

So it did boot the first time. It looks like the chips on the internal memory modules are sometimes different:

I'll now try with the other module to see if the behaviour is the same.

• Same behaviour.

[wessel-novacustom]

On another V560TU with Ultra 7, v0.9.0-rc4 BIOS and belonging EC, 2x 32 GB, 1x PX700 2 TB installed on the MB, the start/boot/restart behaviours are as expected, although booting is a bit slow at about 8/9 seconds till the NovaCustom boot splash screen appears.

[wessel-novacustom]

Tested 6 units so far. 3 of them work fine and these all have an Ultra 7 155H. The 3 failing units have an Ultra 5 125H CPU.

[marmarek]

I forgot to mention mine failing one is Ultra 5 125H too.

[macpijan]

So we are starting to get some pattern emerging at least 🙃

[mkopec]

According to https://review.coreboot.org/c/coreboot/+/81269/2 there's a critical bug in the cache-as-ram logic in the CPU if the cache set count is not a power of 2 :)

On 155H there are 32768 cache sets, on 125H there are 24576. Seems like we're running into this issue, because coreboot hangs just before postcar. Will apply that patch train and see if it fixes anything.

[mkopec]

Yes this fixed the affected V560TU unit with 125H.

[wessel-novacustom]

@mkopec Would be nice if you could prepare another release candidate, maybe also with an iPXE fix.

[mkopec]

-rc5 was sent to @wessel-novacustom @marmarek and the issue is confirmed resolved, I believe.

[wessel-novacustom]

@mkopec I confirm the correct functionality on rc-5, platform V540TU, Ultra 5.