My Lenovo Legion Pro 5 laptop is randomly freezing, completely including ctrl+alt+F3. I have to force power it off and turn it back on.
It often freezes without any message but when a message does appear it is a network disconnect message.
checking the journalctl there is
ACPI BIOS Error (bug). Could not resole symbol [\_SB.PCI0.PB2], AE_NOT_FOUND (20250404/dswload2-162)
I am primarily using the NVIDIA GPU but, have the below error (truncated messages)
amdgpu DMCUB error
bluetooth: Failed to set mode: Failed (0x03)
wpa_supplicant: bgscan simple: Failed to enable signal strength monitoring
I am dual booting Ubuntu and Windows.
Something similar was happening in Windows so I updated drivers, changed some power features, and disabled fast startup. This seemed to fix it. For Ubuntu, I have simply disabled WiFi, using an RJ-45 wire instead, and it seems to be stable now.
While I could try Lenovo support, the laptop is on warranty, this seems like a bad driver issue.
I also tried to updating the UEFI.
Note: Laptop is connecting to a standard Bell Canada Fibe gateway.
What’s the output of journalctl --no-hostname --no-pager --priority=err --boot=-1? -1 is the boot before last, so please adjust accordingly if it was further back.
I may have found a GitHub repository that reportedly fixes the issues: /zbowling/mt7925
It has BETA status for Ubuntu. I would prefer the fix already finalized and in the Linux Kernel (with a simple apt update.
** Edited by Moderators to remove potentially dangerous link.
OK, going by the logs, I think the wifi driver never really comes up, and that makes wpa_supplicant hang, and dominos…
I believe there was a workaround to disable power management in the wifi device (but it was Realtek), perhaps even in the UEFI menu. That could help with the Failed to enable signal strength monitoring, which, I think, is meant for using low-power modes when signals are good. And faulty power management could also cause the device to never enter “power on” state correctly.
First, one should never, ever run such curl ... | $SHELL pipeline! At least no without having at least a cursory look at the script. And second, that installs a package repository and enrolls a GPG key in the APT keyring.
I am sure the author means well, but the implications are plentiful. They could replace any package on your system, without you even realizing it, if you don’t pay close attention.
peter, I read your post but I dont agree. Many apps install via github … | sudo bash
timeshift is available for ubuntu 24 with a simple sudo apt install timeshift. You take a snapshot of your pc, then you install the mt driver. You dont like it, use timeshift to revert back to your previous snapshot of your pc. No repo, no dkms, no mt driver.
You don’t have to agree. I just want everyone to know what they might be in for. You are essentially granting the other party full root access for all eternity by running that script!
First, timeshift is a universe package, so it’s not exactly supported by Ubuntu. And second, timeshift can’t un-compromise a system after the fact. 3rd party repos should be used very sparingly.
Has all this talk, lately, about supply chain attacks still not reached everyone? If I were an evildoer, that’s how I’d spread my root kit. Plus, thanks to DKMS, it’ll be signed with the MOK, so Secure Boot won’t save you either.
We agree to disagree. I use timeshift many times a day, never had an issue. Even if the repo is left there, Sheldon can look in /etc/apt/sources.list.d/ and remove it. Sheldon can apt-key list and remove a key no longer used.
Trust but verify … the github guy is legit, has done great work on wifi drivers.
And how many unsuspecting novice Ubuntu users will be imprinted by with such bad practices? This is a help community and absolute beginners, especially, are to be the expected audience.
But yes, agree to disagree. I’ve said my part on that.
I couldn’t let it go, so I took a gander at that script, and, guess what, Sheldon can’t apt-key list, because that only lists keys in /etc/apt/trusted.gpg.d/; it’s also deprecated, as I’ve just been told by it. The thing is, that key gets “installed” (as in piped) to /usr/share/keyrings/mt76-packages-archive-keyring.gpg.
At the very least, this goes to show that 3rd parties aren’t necessarily good at adhering to established best practices in distro land, in general, and to policies in Debian/Ubuntu, in particular.
Any user installing 3rd party repositories should consider their warranty voided.
Not that there ever was one, but don’t go blaming Canonical/Ubuntu for breaking your toys, if funny stuff starts happening, i.e. after some package upgrades:
Please note, that it’s not the same bug, apparently, but it’s the same MT7925e driver, so you could try with the proposed 6.8 update; don’t know about 6.17, though. If it helps, it may be of interest for said bug report as well.
I agree that installing, especially outside of trust repositories, from GitHub, is a risk. However, I did read over the code on GitHub, as I know it required a MOK install for secure boot to install the third party Kernel level module/driver. On a positive note, the laptop has not frozen yet after the install. This is a huge improvement as, freezing in the middle of work, is worse than a theoretical hacker. Also, when demoing something and the entire computer freezes, it is not a good look. The person who created the repository is Zac Bowling, who is Staff Software Engineer at Meta. His specialty is in Rust language, working on Meta Quest AV/VR (keeping news and politics out of this). Formally worked for Google.
ls /etc/apt/source.list.d/ | grep mt76
ls /usr/share/keyrings/ | grep mt76
ls /etc/apt/sources.list.d/mt76*
searched for, out of paranoia
ls /usr/share/keyrings/mt76*
ls /etc/apt/trusted.gpg.d/mt76*
I have given the install script and had both Gemini and Claude Opus 4.6 scan the repo (along with my own eyes), both do not think that the script somehow imported the key into the archive or main system trust stores.
Risks have to be mitigated. It is better than waiting for an apt update or LTS release to fix a new laptop. The only other alternative was to replace the PCIe card with Intel.