9 Oct 2024 |
emily | you'd already be writing a filesystem driver here. | 19:55:57 |
emily | wouldn't be so bad to make it work with both FUSE and systemd-boot. in the grand scheme of things. | 19:56:09 |
ElvishJerricco | I thought the point was that it was read-compatible with FAT? i.e. No need for a driver for readonly operations? | 19:56:40 |
aloisw | In reply to @emilazy:matrix.org when does firmware write directly to it? Not aware of any particular instances, I was referring systemd-boot like cases. | 19:56:49 |
emily | In reply to @elvishjerricco:matrix.org I thought the point was that it was read-compatible with FAT? i.e. No need for a driver for readonly operations? yeah | 19:57:12 |
emily | but we were just talking about systemd-boot doing writes | 19:57:17 |
aloisw | In reply to @emilazy:matrix.org wouldn't be so bad to make it work with both FUSE and systemd-boot. in the grand scheme of things. Again, systemd-boot does not ship filesystem drivers. | 19:57:23 |
emily | and your Linux distro needs to write | 19:57:26 |
ElvishJerricco | my firmware writes to my ESP when I do a firmware update and it restores my old ubuntu boot loader that I deleted over a year ago 🙃 | 19:57:30 |
emily | In reply to @aloisw:kde.org Again, systemd-boot does not ship filesystem drivers. sure. I mean I'm just figuring out what you can do with existing firmware. | 19:57:40 |
emily | convincing Poettering is another matter, although "reliable FAT32" sounds like the kind of cursed thing he could be sold on. | 19:57:51 |
emily | but at least systemd-boot is something you can patch | 19:58:09 |
ElvishJerricco | In reply to @emilazy:matrix.org and your Linux distro needs to write oh, sure, but I still think we want to avoid including that in systemd-boot | 19:58:20 |
emily | what does systemd-boot write for, that boot counter stuff or whatever? | 19:58:39 |
ElvishJerricco | boot counting and it updates random-seed | 19:58:53 |
ElvishJerricco | Though neither of those things is essential | 19:59:19 |
emily | it feels like some reliability is better than none | 19:59:41 |
aloisw | In reply to @emilazy:matrix.org sure. I mean I'm just figuring out what you can do with existing firmware. You could also write a UDF driver and place it on a tiny read-only FAT ESP. | 19:59:42 |
emily | that also seems reasonable | 20:00:01 |
emily | but is a less fun engineering problem | 20:00:05 |
emily | it also doesn't solve colocating with Windows or whatever (with obvious sacrifice to reliability) | 20:00:24 |
ElvishJerricco | yea, I think the right way to make systemd-boot use a better FS is to have FS drivers in another project | 20:00:25 |
ElvishJerricco | e.g. efifs | 20:00:28 |
emily | why does efifs base its drivers on GRUB rather than, idk, anything maintained? | 20:00:50 |
ElvishJerricco | (efifs specifically is cursed because it's based on grub, but rEFInd also has its own drivers for some linux file systems) | 20:01:00 |
emily | they could do a NetBSD rump kernel or something if the Linux FS interface is too painful | 20:01:04 |
ElvishJerricco | we just need to make LKL capable of being compiled for EFI | 20:01:33 |
ElvishJerricco | /s | 20:01:36 |
ElvishJerricco | (actually wait that sounds sweet) | 20:01:50 |
ElvishJerricco | erofs for the xbootldr partition, anyone? | 20:02:30 |