Re: git: 7919c76dbdd2 - main - bsdinstall: use gpt/efiboot0 label in /etc/fstab
- In reply to: Alan Somers : "Re: git: 7919c76dbdd2 - main - bsdinstall: use gpt/efiboot0 label in /etc/fstab"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Wed, 15 Feb 2023 16:48:42 UTC
On Wed, Feb 15, 2023 at 9:35 AM Alan Somers <asomers@freebsd.org> wrote: > On Wed, Feb 15, 2023 at 2:31 AM Emmanuel Vadot <manu@bidouilliste.com> > wrote: > > > > > > Hello Alan, > > > > On Tue, 24 May 2022 19:17:44 GMT > > Alan Somers <asomers@FreeBSD.org> wrote: > > > > > The branch main has been updated by asomers: > > > > > > URL: > https://cgit.FreeBSD.org/src/commit/?id=7919c76dbdd20161247d1bfb647110d87ca5ee0f > > > > > > commit 7919c76dbdd20161247d1bfb647110d87ca5ee0f > > > Author: Alan Somers <asomers@FreeBSD.org> > > > AuthorDate: 2022-05-24 18:24:12 +0000 > > > Commit: Alan Somers <asomers@FreeBSD.org> > > > CommitDate: 2022-05-24 19:17:10 +0000 > > > > > > bsdinstall: use gpt/efiboot0 label in /etc/fstab > > > > > > Otherwise, boot will hang if the numbering of disks has changed > since > > > initial install. > > > > > > MFC after: 2 weeks > > > Sponsored by: Axcient > > > Reviewed by: brd > > > Differential Revision: https://reviews.freebsd.org/D35309 > > > > So, I understand the initial problem but this causes a new one. > > efibootmgr is now confused when you ask to print the unix path of the > > ESP because the device resolved is still ada0p1 (for example) and since > > it's not mounted anymore as the label is used we can't discover the > > path to the mounted ESP. > > Since it's now in 13.2 we've just noticed that at work. > > We will parse glabel but it's really horrible to parse and it be good > > to have efibootmgr -E -p working again on a fresh install. > > > > Cheers, > > > > -- > > Emmanuel Vadot <manu@bidouilliste.com> <manu@freebsd.org> > > Annoying. Could you please open a bugzilla PR so we don't lose track of > this? > Yea, when I wrote the code I kinda gave up after needing to walk backwards in the geom mesh for places there were no back pointers. I've left more details on the bug that corvind@ filled. Likely a little brute force is all that's needed to find what we need... Warner