RPi400 not starting with ROM 18Mar21

Today’s ROM (2021-03-18), which contains some initial Long Descriptor (8GB etc.) support, fails to start in the RPi400 when powered up from cold. The same SD card works perfectly in the 8GB RPi4. A Fat32 USB stick with today’s ROM also fails to start in the RPi400.

There are no signs of life from the stalled RPi400, no VNC output is seen

To get the RPi400 to start an earlier ROM is required, then after re-installing today’s ROM the RPi400 will restart from the TaskManager icon.

There has been no such problem with previous ROMs.

Firmware is 05Mar21 which is OK with earlier ROMs and in the 8GB RPi4.

(P.S. Now also tested with firmware 15Mar21.)

The RPi400 may need to be switched off for about 5 minutes to reproduce this.

(Today’s ROM is also OK in the Titanium.)

The RPi400 may need to be switched off for about 5 minutes to reproduce this.

That suggests that maybe the problem’s been caused by removing the RAM clear code from the HAL – i.e. there’s some key bit of memory which the isn’t being initialised properly on startup. If you have the time, could you try a local build which has the previous version of the HAL? (HAL_BCM2835-0_93). If not, I can upload a couple of builds later tonight (and do some more testing to see if any of my Pis fail – hopefully my 4GB Pi will be a suitable stand-in for a Pi400)

could you try a local build which has the previous version of the HAL? (HAL_BCM2835-0_93)

A build with HAL_BCM2835-0_93 is fine, the RPi400 is starting perfectly.

hopefully my 4GB Pi will be a suitable stand-in for a Pi400

I don’t have a 4GB RPi4 so cannot confirm that.

Pleased to do further testing.

I have a Pi4 8Gb and have no issues with ROM 18-Mar-2021

The hdmi_blanking=1 command was removed from config/txt to see if the stall could be seen in the ROM initialisation messages. The startup does not get that far. The monitor indicator changes from amber to green, syncs and black level, but that is all.

The RPi400 had been overclocked to 2400MHz which had been fine. All overclocking was removed, just in case, but the startup still fails.

The odd bit is that the RPi400 needs to have been off for a while, five to ten minutes, for the fault to show.

All of my Pis appear to be fine. I’ll see if I can provoke a failure by filling RAM with garbage on startup.

filling RAM with garbage on startup.

Since when did we have a Facebook app?

Since when did we have a Facebook app?

LOL as the saying goes.
I read it out to the wife, and… LOL much.

I think Jeffrey’s building one from the ground garbage up.

All of my Pis appear to be fine. I’ll see if I can provoke a failure by filling RAM with garbage on startup.

Happens here on a Rpi 400 on power on. Booting off an SD card with a ROM before the 18th then works again without powering off. Reverting to the new ROM (18th Mar) and rebooting also boots up fine.

Filling RAM with garbage seemed to do the trick – there was a bit of kernel workspace that wasn’t being initialised properly.

Here are a few test ROMs: http://www.phlamethrower.co.uk/misc2/bcm2835dev.zip

• rURZ00-00 is a normal ROM – please check that this one does fail on the Pi400 (just to make sure some difference in build setup doesn’t mean the other test results are invalid)
• sURZ00-00 has my fixes applied, so it should work
• tURZ00-00 has my fixes and the code to fill RAM with junk; if the ‘s’ ROM fails then it would be good to try this one and see if the behaviour is any different

rURZ00-00 is a normal ROM – please check that this one does fail on the

From power on this one fails to boot.

sURZ00-00 has my fixes applied, so it should work

From power on this one has successfully booted every time.

I find the same, rURZ00-00 fails and sURZ00-00 is good.

Cool; thanks for the help guys. Merge request pending.

https://gitlab.riscosopen.org/RiscOS/Sources/Kernel/-/merge_requests/52