Raspberry Pi 4 usb boot?

[jamesh]

Yes, network boot has been added.

Is there any documentation available regarding network boot on RPi-4? I've trawled the board but most of what I've found looks to be a series of solutions to encountered problems …

TIA

Still going through merging, should be up soon. See https://github.com/raspberrypi/rpi-eepr ... ot_beta.md

[harwoodr]

Still no movement on booting RPi4 from usb....they dont seem to realise that to some users boot speed and reliabilty are important, come on guys you promised that this feature would be added, if it doesnt come soon we will be into Rpi 5 already

[jj_0]

Still no movement on booting RPi4 from usb....they dont seem to realise that to some users boot speed and reliabilty are important, come on guys you promised that this feature would be added, if it doesnt come soon we will be into Rpi 5 already

I really don't understand this kind of comment. As has been stated many times in this thread (I guess it's my turn now) you can already boot with your rootfs on USB with only the initial boot files on an SD-card. There is really no speed nor reliability advantage against booting entirely from USB. The only advantage I see is that booting entirely from USB frees up the SD-card slot.

Patience is a virtue.

[RonR]

The only advantage I see is that booting entirely from USB frees up the SD-card slot.

And if you really need the SD card slot for some other use, you can remove the SD card that's used for booting as long as you replace it before trying to boot to USB again.

[rpdom]

As has been stated many times in this thread (I guess it's my turn now) you can already boot with your rootfs on USB with only the initial boot files on an SD-card. There is really no speed nor reliability advantage against booting entirely from USB.

Exactly the opposite, in fact. The Pi 3B[+] for example will try to look for an SD card to boot from for a few seconds before it tries to boot from USB or network. So the start of the boot at least will take longer than just booting initially from SD card.

[jamesh]

Still no movement on booting RPi4 from usb....they dont seem to realise that to some users boot speed and reliabilty are important, come on guys you promised that this feature would be added, if it doesnt come soon we will be into Rpi 5 already

Q. Do you think we are stupid
A. No, we are not, and someone is working on it right now. You made a colossal mistake - that of think that just because YOU haven't seen any progress, that there hasn't been any progress. News for you, that's not how it works.

You clearly do not realise how complex it actually is to do this. Requires an entire open source (not GPL) USB 2/3stack to be found/developed, to cope with whatever can be plugged in. Plus a USB HW driver that works via the PCIe interface, so PCIe also needs to work. All in the bootloader! Huge amount of very complex work.

As for boot speed - already referenced above - quicker to use an SD card!!!!!

Reliability? Got any figures? I've had terrible reliability from SSD's and great reliability from SD cards...

[clicky]

Still no movement on booting RPi4 from usb....they dont seem to realise that to some users boot speed and reliabilty are important, come on guys you promised that this feature would be added, if it doesnt come soon we will be into Rpi 5 already

Q. Do you think we are stupid
A. No, we are not, and someone is working on it right now. You made a colossal mistake - that of think that just because YOU haven't seen any progress, that there hasn't been any progress. News for you, that's not how it works.

You clearly do not realise how complex it actually is to do this. Requires an entire open source (not GPL) USB 2/3stack to be found/developed, to cope with whatever can be plugged in. Plus a USB HW driver that works via the PCIe interface, so PCIe also needs to work. All in the bootloader! Huge amount of very complex work.

As for boot speed - already referenced above - quicker to use an SD card!!!!!

Reliability? Got any figures? I've had terrible reliability from SSD's and great reliability from SD cards...

This was quite unnecessary. Is it called 'feeding the troll''?

Nobody with half a brain and good intentions do not presume that you're not doing anything. On the contrary! How many times we were positively surprised with new developments (I'm just referring to everything you've done after original Raspberry Pi was released)?

People who are using Raspberry Pi 4 B daily (like me - booting it from SSD) do not complain nor have any issues at all. People who refuse to turn on 'common sense' part of the brain will complain about pretty much everything. Just imagine what kind of barrage of 'negativity' will spawn from introducing USB boot - it will be by default broken, not worth, bad, and who knows why not suitable (for them)...

[LTolledo]

As a temporary measure am also booting my RPi4B-4G with M.2 SSD using RonR's script to set it up...

no real problems so far.....

[pepeEL]

As a temporary measure am also booting my RPi4B-4G with M.2 SSD using RonR's script to set it up...

no real problems so far.....

Please share this script with manual how do it?

[jamesh]

Still no movement on booting RPi4 from usb....they dont seem to realise that to some users boot speed and reliabilty are important, come on guys you promised that this feature would be added, if it doesnt come soon we will be into Rpi 5 already

Q. Do you think we are stupid
A. No, we are not, and someone is working on it right now. You made a colossal mistake - that of think that just because YOU haven't seen any progress, that there hasn't been any progress. News for you, that's not how it works.

You clearly do not realise how complex it actually is to do this. Requires an entire open source (not GPL) USB 2/3stack to be found/developed, to cope with whatever can be plugged in. Plus a USB HW driver that works via the PCIe interface, so PCIe also needs to work. All in the bootloader! Huge amount of very complex work.

As for boot speed - already referenced above - quicker to use an SD card!!!!!

Reliability? Got any figures? I've had terrible reliability from SSD's and great reliability from SD cards...

This was quite unnecessary. Is it called 'feeding the troll''?

Nobody with half a brain and good intentions do not presume that you're not doing anything. On the contrary! How many times we were positively surprised with new developments (I'm just referring to everything you've done after original Raspberry Pi was released)?

People who are using Raspberry Pi 4 B daily (like me - booting it from SSD) do not complain nor have any issues at all. People who refuse to turn on 'common sense' part of the brain will complain about pretty much everything. Just imagine what kind of barrage of 'negativity' will spawn from introducing USB boot - it will be by default broken, not worth, bad, and who knows why not suitable (for them)...

Entirely necessary.

[thagrol]

As a temporary measure am also booting my RPi4B-4G with M.2 SSD using RonR's script to set it up...

no real problems so far.....

Please share this script with manual how do it?

I think you'll find it has been linked to at least once in this thread already. You did read the thread before posting didn't you?

Failing that a forum search shold find it (hint: try "usb boot made easy")

[B.Goode]

As a temporary measure am also booting my RPi4B-4G with M.2 SSD using RonR's script to set it up...

no real problems so far.....

Please share this script with manual how do it?

If the answer given the last time you asked in this thread is not adequate, perhaps it would be more productive to say in what way it does not meet your expectations?

If there is some restriction imposed by the third-party Operating System you hope to use, maybe ask the developer/maintainer of that release to help you....

https://www.raspberrypi.org/forums/view ... 4#p1550329

[jamesh]

Q. Do you think we are stupid
A. No, we are not, and someone is working on it right now. You made a colossal mistake - that of think that just because YOU haven't seen any progress, that there hasn't been any progress. News for you, that's not how it works.

You clearly do not realise how complex it actually is to do this. Requires an entire open source (not GPL) USB 2/3stack to be found/developed, to cope with whatever can be plugged in. Plus a USB HW driver that works via the PCIe interface, so PCIe also needs to work. All in the bootloader! Huge amount of very complex work.

As for boot speed - already referenced above - quicker to use an SD card!!!!!

Reliability? Got any figures? I've had terrible reliability from SSD's and great reliability from SD cards...

This was quite unnecessary. Is it called 'feeding the troll''?

Nobody with half a brain and good intentions do not presume that you're not doing anything. On the contrary! How many times we were positively surprised with new developments (I'm just referring to everything you've done after original Raspberry Pi was released)?

People who are using Raspberry Pi 4 B daily (like me - booting it from SSD) do not complain nor have any issues at all. People who refuse to turn on 'common sense' part of the brain will complain about pretty much everything. Just imagine what kind of barrage of 'negativity' will spawn from introducing USB boot - it will be by default broken, not worth, bad, and who knows why not suitable (for them)...

For those who think I'm being a bit harsh, here's an analysis.

"Still no movement on booting RPi4 from usb" Untrue statement, work is being done.

"....they dont seem to realise that to some users boot speed and reliability are important" Yes, we do. Some users. Not a huge percentage btw. And incorrect in that SD card/USB boot is faster, and unduly pessimistic about SD card reliability, which is fine.

"come on guys you promised that this feature would be added" Yes we did, and it will be. We said it would be months, and it's taking months. We are doing exactly what we said we would do.

"if it doesnt come soon we will be into Rpi 5 already" No, it really won't. Pi5 is years away. USB cardless boot is months away.

[k-pi]

They released early so that manufacturers could get a head start, & that's likely the reason the 4GB version was in short supply.
(A certain manufacturer of thin clients had a supply).

This RPi4 version wasn't really due to be released until 2020, & I'm guessing that is why so much of the software drivers isn't ready.

I've put mine aside until all of it functions work as promised, it's an unfinished product, it will hopefully be fully functional eventually, but at this rate, not likely before next February, at a guess.

[dickon]

You're just trolling now.

[jamesh]

They released early so that manufacturers could get a head start, & that's likely the reason the 4GB version was in short supply.
(A certain manufacturer of thin clients had a supply).

This RPi4 version wasn't really due to be released until 2020, & I'm guessing that is why so much of the software drivers isn't ready.

I've put mine aside until all of it functions work as promised, it's an unfinished product, it will hopefully be fully functional eventually, but at this rate, not likely before next February, at a guess.

All complete rubbish. One more like that are you are out of here. It's concern trolling pure and simple. In fact, no, I think you are going right now. You've been warned before.

[jcyr]

You clearly do not realise how complex it actually is to do this. Requires an entire open source (not GPL) USB 2/3stack to be found/developed, to cope with whatever can be plugged in. Plus a USB HW driver that works via the PCIe interface, so PCIe also needs to work.

Sounds like a lot to stuff into the EEPROM? Too bad you can't use GPL code... Is this a consequence of the VC code being closed source proprietary code?

[W. H. Heydt]

This RPi4 version wasn't really due to be released until 2020, & I'm guessing that is why so much of the software drivers isn't ready.

That was a function of the expected number of revision cycles for the silicon. Since an earlier than expected version worked, the release was moved up and many, many people are happily using Pi4B boards as they are.

I've put mine aside until all of it functions work as promised, it's an unfinished product, it will hopefully be fully functional eventually, but at this rate, not likely before next February, at a guess.

The more fool you.

[jamesh]

You clearly do not realise how complex it actually is to do this. Requires an entire open source (not GPL) USB 2/3stack to be found/developed, to cope with whatever can be plugged in. Plus a USB HW driver that works via the PCIe interface, so PCIe also needs to work.

Sounds like a lot to stuff into the EEPROM? Too bad you can't use GPL code... Is this a consequence of the VC code being closed source proprietary code?

Not exactly, somewhat more to do with bootloader security AIUI, but could be wrong, and of course, if one bit is GPL, it all needs to be GPL. And we would then need to supply a compiler or provide the means to get hold of it....and we are not allowed to distribute the compiler - it's not ours to do so. And it is not available to the general public from a third party.

So, a number of reasons.

More information here https://www.softwarefreedom.org/resourc ... guide.html

[RonR]

As a temporary measure am also booting my RPi4B-4G with M.2 SSD using RonR's script to set it up...

no real problems so far.....

Please share this script with manual how do it?

Running Raspbian from USB Devices : Made Easy

[RossDv8]

I was running my Pi 4 from a 128GB Samsung EVO Plus microSd card and planning to re-image my 240GB SSD that had been used on the Pi 3B+ later when the USB boot had been sorted. But I noticed a few performance issues with the microSD, and decided to try it sooner.
So yesterday I followed the old instructions, and imaged the SSD using etcher, then imaged an old Sandisk ultra 16GB card that was not being used and followed the directions to modify the cmdline.txt - and it didn't work.

I had ended up with something like a 7GB SSD, but I figured I could ignore the unallocated space and do a resize later with gparted or something, then once it was booting (initiaalised from the microSD) I would just copy all my /home/pi stuff over..
I tried a couple of other things which also failed. Then I read a supposedly 'simple' method, and used SD Card Copied from the Accessories menu.

Once I modified the UUID for the root file system, the thing booted immediately and lo and behold, my personal desktop layout and all the icons for my shell scripts were in their panel.

So I opened the file manager, expecting to see the SSD listed as 128GB total, since it was cloned from the microSD card.
I was pleasantly surprised to see Free Space of about 192GB and Total Space of 218GB, which is about right because the WD 240GB SSD always reports 220GB. My guess is there might be some sort of redundant space for maintenance or whatever.

Anyway, no need to resize anything. And everything has been working brilliantly.
So for anyone holding off setting up their SSD because they are waiting for firmware or something to change - I reckon it's probably not worth the wait, because I doubt the start time to a usable screen is more than 2 seconds different.

Performance seems about average, but considerably faster than the microSD.

pi@raspberrypi:~ $ sudo hdparm -tT /dev/sda

/dev/sda:
Timing cached reads: 1542 MB in 2.00 seconds = 770.55 MB/sec
Timing buffered disk reads: 874 MB in 3.00 seconds = 290.88 MB/sec
pi@raspberrypi:~ $

Which was great until I read this post:
https://www.raspberrypi.org/forums/view ... d#p1488304

Where it said:
'... if needed), and edit /etc/fstab to mount the SD card as /boot, so that kernel and firmware updates are properly applied.'

So I commented out the line with the SSD /boot UUID on it, and replaced it with the UUID of /boot on the SD card. Saved ad did a reboot.
And got a Windows Blue Screen Of Death! Well, that's what it looked like. Powered off for a few seconds and powered on, and the Pi was up and running. Tried it a few times with the same result.

So I edited /etc/fstab and removed the new line, and the # in front of the old one. Tried reboot, and same problem.

Now, I only changed one line from this:

proc /proc proc defaults 0 0
PARTUUID=b2fbcf8a-01 /boot vfat defaults 0 2
PARTUUID=b2fbcf8a-02 / ext4 defaults,noatime 0 1
# a swapfile is not a swap partition, no line here
# use dphys-swapfile swap[on|off] for that

to This:

proc /proc proc defaults 0 0
#PARTUUID=b2fbcf8a-01 /boot vfat defaults 0 2
PARTUUID=5e3da3da-01 /boot vfat defaults 0 2
PARTUUID=b2fbcf8a-02 / ext4 defaults,noatime 0 1
# a swapfile is not a swap partition, no line here
# use dphys-swapfile swap[on|off] for that

5e3da3da-01 is /boot on the microSD
b2fbcf8a-01 is /boot on the SSD
b2fbcf8a-02 is root on the SSD

I would have thought reverting to the original fstab should return my reboot function to normal.

If someone knows a reason it doesn't, please let me know.
If someone can confirm that next time I do this I should NOT edit fstab so the /boot of the microSD card is mounted, that would be great too.

It was working quite nicely using /boot on the SSD
For now, until I decide to reimage the SSD I will just use the power switch in to reboot.

[HawaiianPi]

You're just trolling now.

That's mostly what k-pi does. It's gotten even worse since the release of the Pi4, but he used to troll Pi3 USB boot threads as well.

I'm fine with /boot on SD for now, and would rather have that than wait for the Pi4 with USB boot later.

[jamesh]

You're just trolling now.

That's mostly what k-pi does. It's gotten even worse since the release of the Pi4, but he used to troll Pi3 USB boot threads as well.

I'm fine with /boot on SD for now, and would rather have that than wait for the Pi4 with USB boot later.

Dealt with, no need for further comment.

[DougieLawson]

Don't move /boot to the SSD until it's being used to boot the system. While it's booting from the (small) SDCard you need to have that partition mounted at /boot or it WILL break when the raspberrypi-kernel and raspberrypi-bootloader packages get an update.

[RossDv8]

Don't move /boot to the SSD until it's being used to boot the system. While it's booting from the (small) SDCard you need to have that partition mounted at /boot or it WILL break when the raspberrypi-kernel and raspberrypi-bootloader packages get an update.

Question to DougieLawson or perhaps jamesh

I'm curious. When the SSD is imaged, fstab points to /boot and /root as BOTH being on the SSD.
I read the post that said to edit fstab, as well as your reply above, to mean that fstab should be changed as I showed in my post.
But when I had changed it, reboot no longer worked.

So why should fstab be changed to show a reference to /boot on the microSD, rather than the default of /boot on the SSD?
From wat I could make out, leaving fstab showiing /boot on the SSD when the initial boot is done from the microSD can screw up kernel upgrades, firmware etc.
A reminder that I did not 'Move' /boot to the SSD. I edited the line in fstab to 'point' to /boor on the microSD.- as per someone's iinstruction that not doing the edit would mess up kernel and firmware updates.

But if it 'should' be changed - why does that change kill reboot?