Re: Amdgpu 15.1
"Edward Sanford Sutton, III" <[email protected]> Mon, 3 Aug 2026 12:17:04 -0700
| Newsgroups | gmane.os.freebsd.questions |
|---|---|
| Message-ID | <SA1PR11MB881155E82C3E8DBF807C31B7E6D52@SA1PR11MB8811.namprd11.prod.outlook.com> |
On 8/2/26 13:37, Gerard A wrote: > The error message is after updating it panics because the amd drivers > are now broken Was this 15.1-RELEASE-p1 -> 15.1-RELEASE-p2 or were you upgrading from a different version? > But the real issue is that after trying to remove the amdgpu shit from > loader.conf rc.conf amdgpu should only be in rc.conf and not loader.conf. > And running freebsd-update rollback > > My zfs pool is no longer mountable > No longer visible with zfs import and > > zdb, pointed at the partitions, returns. > Failed to unpack label 0 / 1 / 2 / 3 / 4? I presume you didn't get that far due to amdgpu panic but did you take any manual steps to update the zpool after upgrade? Do you have ZFS boot environments? Running `bectl list` should display entries if so and the boot loader should offer to pick different entries. I haven't used `freebsd-update rollback` but presume that shouldn't happen. If no bug report already exists for it and the issue doesn't point to something else then it would likely be very good to make sure a problem report is created and tracked specifically for this. > So its totally fucked now for reasons that are not clear at all... How > does that happen to both disks of RAID1 RAID1 does not offer any protection against software error and user error; that is what backups, and less so zpool checkpoints, and even less so zfs snapshots, and least so boot environments help against. If a drive failed or its data was corrupted then RAID1 helps but when the user or program performs any change to the data it goes to both disks. An update that makes amdgpu unusable goes to both disks. Any freebsd-update related failure goes to both disks. UEFI and older BIOS based boot loaders are written to disk outside of ZFS. I don't use freebsd-update but I thought it was still left up to the user to update the boot loader separately and that it didn't execute any ZFS commands other than stuff to create boot environments; I don't know if rollback will delete created boot environments but if you weren't actively doing anything with them then you should have still been on one called 'default' both before and after upgrade with the only question being if it rollback deleted or picked the previously created boot environment but I'd doubt it does either. If you don't want an activity to impact both disks, you can split or offline one of the two disks but while separated you will lose the protections that RAID normally offers your data. > On 3 August 2026 6:08:52 am AEST, Paul Beard <[email protected]> wrote: > > The error messages and diagnostic information you meant to send > seems to have gotten lost. Please try resending that and maybe > somebody can help you. > -- > This space intentionally left blank. > > On Aug 2, 2026, at 12:58 PM, Gerard A > <[email protected]> wrote: > > > Love how i had a working system and now i don't because i > updated.... For fucks sake cunts > >