Pi 4 Boots up correctly on 2nd attempt

My Pi 4 has started booting up from cold without running !Boot, but on shutting down from the iconbar cog menu followed by a restart – so not from cold – it boots up correctly. It’s running RISC OS 5.30 if that helps.

It may be relevant that !Boot contains the DOSDisc, Loader, and usually runs with a USB stick containing the Linux stuff where EDOS deals with switching between the two OSs.

The problem has occurred only yesterday and I haven’t changed anything in the configuration or settings.

Your power brick may be on the way out. Try using another power brick: ideally you’d want one designed to work with a Pi 4, but a USB-C mobile phone charger might do as a proof-of-concept (provided that it’s rated at 5 volts 3 amps).

I have found a mobile phone charger in the house with output rated as 5V 3A/ 9V 2A. There are no switches or buttons on the adaptor to choose between the two outputs, and so, presumably it can detect the appropriate voltage?

Yep, USB uses some sort of “negotiation” process to figure out the appropriate voltage.

Yes, it will default to 5V until some complex negotiation gets it to step up to something higher.
Edit: Chris got in ahead of me because I got stuck on a wiki walk reading junk about charging protocols (Power Delivery) that I really didn’t need to know… 🥴

I have just tried the Pi 4 again with the mobile phone charger, but again it boots up without reading !Boot. So there must be something else going on.

Could it be *opt 4,2?

EDOS writes to the Loader partition to alter (IIRC) the CONFIG.TXT file to change which system boots next. You need to check the contents of CONFIG.TXT on both SCSI (USB) and SDFS and the contents of CMOS on both to check the correct boot filesystem is specified.

I prefer a push button to select boot file system (as written up in Archive 25:5.

I prefer a push button to select boot file system (as written up in Archive 25:5.

I ought to do something on my blog about that. Something accessible.

@Chris Hall
On the SD card (SDFS) the only line in the CMOS file that seems relevant to me is line 5: C0 Filing system number. The config/txt file on the SD card is dated 14:27:50 05 DEC 2021, and so hasn’t been changed.

On the USB stick (SCSI) the CMOS file is identical to the one on the SD card, at least for the first 240 lines (the length to which Stuart Painting’s CMOS-decode goes). There is no config.txt file; but it has a file config.lin which is also in !Boot.Loader, both the same, with the same dates in May 2021.

The config/txt file on the SD card has the line kernel=RISCOS.img .

I ought to do something on my blog about that. Something accessible.

Perhaps something like this

EDOS renames the files CONFIG.LIN to CONFIG.TXT and CONFIG.TXT to CONFIG.ROS (I think) to switch between operating systems – the two versions of CONFIG.TXT therefore are each set up to load the relevant OS. So it does not update the file datestamp.

My RPi 4 has just booted correctly from power on. This is after 3 days of it needing a shut down and restart; but I haven’t changed anything! I don’t know whether this will continue yet.

In case it’s not clear, my EDOS set up powers up into RISC OS, not Linux.

@Chris Hall
I couldn’t find the file you call CONFIG.ROS or any other config files other than config.txt and config.lin; but I did find the line
“sudo mv /mnt/Loader/config.ro5 /mnt/Loader/config.txt”
in the /Reboot file, which is odd.

A question: On powering up from cold, does the hardware look at the SD card or the USB stick first?

On power up from cold the Pi 4 EEPROM loads start4.elf and looks at CONFIG.TXT and CMDLINE.TXT to decide what OS to start up. On your set up that starts RISC OS.

If you start Linux from RISC OS (or tell RISC OS that you want Linux to boot on the next boot, not sure which) then it will move or rename the file config.lin to config.txt so that config.txt will have instructions to start Linux. When Linux reboots it then puts things back as they were config.ro5 moves to config.txt so that it now has instructions to start RISC OS on next boot.

On rebooting from RISC OS to RISC OS it updates CMOS then starts us the same (I think) as from cold.

As there is no real time clock on the CM4 the confusion between how Linux stores the date and how RISC OS stores the date (an exact number of decades different) does not arise as they each get their time from the network.

My favoured set up is as shown below:

chris@raspberrypi:~ $ lsblk
NAME         MAJ:MIN RM   SIZE RO TYPE MOUNTPOINT
sda            8:0    1  14.5G  0 disk 
└─sda1         8:1    1  14.5G  0 part /media/chris/USB00
mmcblk0      179:0    0   7.3G  0 disk 
├─mmcblk0p1  179:1    0   300M  0 part /media/chris/EMMCBOOT
└─mmcblk0p4  179:4    0     7G  0 part 
mmcblk0boot0 179:32   0     4M  1 disk 
mmcblk0boot1 179:64   0     4M  1 disk 
nvme0n1      259:0    0 953.9G  0 disk 
├─nvme0n1p1  259:1    0   768G  0 part 
├─nvme0n1p2  259:2    0   300M  0 part /boot
└─nvme0n1p3  259:3    0 185.6G  0 part /
chris@raspberrypi:~ $ 

That is an eMMc containing the standard two-partition RISC OS image, a FAT formatted USB drive to share files between RISC OS and Linux and a 4k sector formatted NVMe drive with a 768GB RISC OS partition and a 300MB FAT and a 185GB ext4 Linux partition. RISC OS cannot see the Linux partitions and Linux cannot read the RISC OS partition.

My Pi 4 has booted up correctly from power on again this morning, and so I am still puzzled as to why it should now do this after many times consistently needing a second attempt via a restart to read !Boot. I haven’t changed anything.

I still haven’t found the file config.ro5 or config.ros. So where is it?

Is it a little cooler at the moment in the room where the Pi4 is, David?

Probably, but not by much. I am in Surrey, not Scotland!
I was wondering about some physical cause, as the problem started just after the Pi 4 had been transported over 200 miles in a case by public transport. But physical jolting doesn’t explain this particular misbehaving. What effect could temperature have?

But physical jolting doesn’t explain this particular misbehaving. What effect could temperature have?

Both of them, separately and collectively, can induce “dry joint” situations for components on a board.
There’s also another factor in the 200-mile shift – peak voltage of the grid supply in the area of use (240V is a nominal value rather than an absolute)

Since harmonisation its 230V -6%/+10% not 240V anymore.

That’s the nominal voltage. There are plenty of older transformers around still chucking out 240V, and since -6/+10% is around 218V-253V, it’s perfectly within range and acceptable.

Here, my power used to vary between 225V or around 212V when I put the kettle on (sometimes as low as 198V), but since the changeover to smart meters, which required a massive overhaul of the supply network, I’m getting a nice ~230V, even with the kettle on.

It’s easy to sign a piece of paper saying “we’ll all agree to 230V”, but it’s something else to change all of the transformers to ensure that 230V is what actually arrives at the socket. The tolerances are probably so that there was plenty of headroom for the 220V/240V equipment, which will be upgraded as needed.
It’s worth observing that 230V plus 10% is not at all far from 240V plus 6%; so the upper limit may well be higher specifically to allow for the UK.

At least it’s not America’s silly little plugs and dweeby 110V. I still can’t wrap my head around Alec of Technology Connections boiling water in a saucepan. That just seems so very strange, until I remember (or he gleefully points out) that electric kettles aren’t really a thing because the usual household supply isn’t up to doing it the European way. ¹

Hmm, if I remember correctly, half the voltage would require twice the amps. So a regular 2kW kettle is around 8.7A at 230V, but a little over 18A at 110V. Yikes.

¹ Yes, I’m aware of the T shaped prong for 20A sockets and that, due to their arcane method of getting that 110V, there’s also a “sort of” 240V available for those more demanding appliances, but since that goes phase to phase rather than the expected phase to neutral, there may be safety considerations preventing one from simply importing a kettle from Europe, shoving an NMEA plug onto it, and running it from 240V.
Of course, this is assuming you even have all of this stuff. Older properties may have straight 15A sockets and kitchen equipment hardwired to the fusebox.

The two places I use this R-Pi 4 are Reigate, Surrey and Ilkley, Yorkshire over several years, and this is the first time I have seen this problem. So the voltage hypothesis seems unlikely.

I’m a connections design engineer for a DNO, most transformers ones prior to harmonisation are rated to 433V nominal at 3ph, 480V nominal at split and 250V nominal at 1ph.

Then tapped accordingly to ensure that the bars high and low are within statutory limits.

Newer ones are rated 415V and 240V then tapped accordingly.

But with the advent of PV, we’ve started not altering the tap position at LV, we do it on the 6.6/11/20kV network to control voltage output as these are onload tap changers at primary substations opposed to offload that are on distribution substations

Electric kettles seem to be about 1200w 110v in the USA – you seem to have to wait longer for your tea :-)

There’s also another factor in the 200-mile shift – peak voltage of the grid supply in the area of use (240V is a nominal value rather than an absolute)

Have a look at the rating label of the PSU in use. The chances are that it’s “universal” input, i.e. good for anything between about 100V and 250V, so the local mains voltage is very unlikely to be a factor unless the particular specimen is very faulty.

If the power supplied to the computer is an issue, it’s more likely to be caused by the cable between PSU and computer being too long and/or thin, so the 5V (or whatever) launched into the cable is significantly less at the destination.