[SOLVED] Boot errors on XPS 15

[frisket]

I am trying to get Linux — any Linux — to install bootably from USB on a DELL XPS 15 laptop (note I am not trying to dual-boot with Windows; I just want Linux). Most distros I have tried (all the *Ubuntus, Arch, RedHat, Fedora) fail even to boot from the USB, hanging irretrievably while their logo spins. Booting in text-mode reveals a known problem ending with the error message

Code: Select all

BUG: soft lockup: CPU#8 stuck for 23s!

(I have been unable to determine if this is a real CPU problem, or something masking the known NVidia errors). Bear in mind that this machine has SSD, not a hard disk.

However, two distros do boot from USB: Bodhi and Mint. Bodhi even installs and can be booted afterwards, but it hangs when you shut down (after reaching target shutdown; it fails to power off), and I want a much more recent version of Enlightenment than the one they provide, which is hard to install because of library conflicts.

So it's Mint, which installs, but when I try to reboot, I just get a black screen and a blinking cursor: no Grub, so no boot. I followed very good instructions from https://community.linuxmint.com/tutorial/view/245 about how to boot from USB and mount the newly-installed system and then force a grub-install to the ext4 root partition, which on an SSD is /dev/nvme0n1p1. I tried this after installing in UEFI mode (failed) and then in BIOS Legacy mode, which seems to have worked, as I now get a Grub screen. But when it starts to boot, it dumps core while still in non-graphical TTY mode, and I get a raft of stuff shoot off the top of the screen, so I now have 23 lines mostly all like this

Code: Select all

[    1.118968]  [<ffffffffaf29d1a2>] rcu_process_callbacks+0x43a/0x5c0

(with different PC values, functions, and offsets). One line actually says

Code: Select all

[    1.119135]  <EOI>  [<ffffffffaeb9e7bf>] ? panic+0x1e6/0x226

and the bottom line says

Code: Select all

[    1.119855] ---[ end trace 25e43f56303d5ee7 ]---

Clearly this is doubleplusungood. As the start of the dump vanishes off the top of the screen in sub-second time, and there don't appear to be any working control keys to scroll back, I can't see if there was anything important there to report. I tried adding nomodeset xforcevesa to the boot line in Grub, but it gets the same coredump, so I am assuming at this stage that it is not a graphics driver error.

It feels like I'm close, but the coredump is beyond my abilities to act on. It's like Grub installed, but the boot process is being pointed to some place out in the void from which it can't recover. Is there some special action I should take while doing grub-install that understands SSDs, for example?

[WharfRat]

Hello frisket

Welcome to Linux Mint and the Linux Mint forum

Since you cannot boot any linux distro it sounds like an overclocking issue or suspiciously like a bad PSU as found here

[michael louwe]

@ frisket, .......

...

.
Try the fix for Nvidia graphics card ... https://linuxmint.com/rel_sonya_cinnamon.php

Ensure that the BIOS setting for the internal hard-drive is AHCI and not RAID; Fast Boot and Secure Boot are disabled. If Win 10 is still present on the internal SSD, Fast Startup in Win 10 has to be disabled.(>Control Panel, >Power options)

Better and easier(less complicated) to install LM in Legacy BIOS mode(rather than UEFI mode) by enabling CSM or Legacy BIOS in BIOS setup.

[frisket]

On Thu Dec 07, 2017 8:46 pm, WharfRat wrote

Welcome to Linux Mint and the Linux Mint forum

Thank you.

Since you cannot boot any linux distro

But I can — both Bodhi and Mint boot from USB, and Bodhi even boots from the SSD after installation.

it sounds like an overclocking issue or suspiciously like a bad PSU as found here

I'm not sure I understand why it would be overclocked: it's new out of the box.
But I did wonder about it actually having a hardware problem — it certainly shouldn't, but I'll check it out.
The Windows supplied preinstalled, and the Windows corporate image my colleague installed, both appeared to work fine, although that may not mean anything.

On Fri Dec 08, 2017 3:22 am, michael louwe wrote:

Try the fix for Nvidia graphics card ... https://linuxmint.com/rel_sonya_cinnamon.php

I did something similar during my trials with Bodhi, but I think I'm past that: the problem is occurring long before the system gets around to worrying about graphics cards.

Ensure that the BIOS setting for the internal hard-drive is AHCI and not RAID; Fast Boot and Secure Boot are disabled.

It's an SSD, not a HDD but I'll check the BIOS. Fast and Secure are already disabled.

If Win 10 is still present on the internal SSD, Fast Startup in Win 10 has to be disabled.

Windows is long gone: I have no use for it, so all my tests have been done as whole-disk installs, allowing Linux to repartition and reformat as it wanted by default.

Better and easier(less complicated) to install LM in Legacy BIOS mode(rather than UEFI mode) by enabling CSM or Legacy BIOS in BIOS setup.

That's what I have done. I tried it once in UEFI mode but it didn't work, so everything since has been in Legacy mode.

P

[frisket]

I think I may have found the solution (now using 18.3 instead of 18.2):

1. Don't use unetbootin to create a Mint USB, use dd instead. Apparently unetbootin does something odd with Mint while writing the USB.
2. Install Cinnamon, not Maté. It appears that Cinnamon requires higher-spec hardware, so it checks for it, and therefore finds problems Maté misses.

The result of this is that the USB booted correctly, but displayed a warning that it was running without hardware support. This did not appear on any of my previous attempts, which is why I am deducing that it is checking something not previously checked for.

It installed to the SSD without error (earlier attempts installed, but issued a warning that it was unable to set up the repos for the CD correctly (!), and then failed to create a boot partition and install Grub. That in itself probably didn't matter (just looked embarrassing) but I noticed that it did install Grub explicitly and do a grub-update; again this was either missing from previous installs, or wasn't being flagged.

Rebooting then worked: previously the system hung on shutdown and needed the power button. Now it starts up correctly, and appears all to be in running order.

I don't know if this was some minor update from 18.2 to 18.3 just happened to fix the specific error that the XPS 15 was creating, but if so, my grateful thanks to whoever on the dev team did it.

Thank you to all those who made suggestions. Now I just need to work out how to mark this one SOLVED.