Arch Linux User Repository

Do this package perform a nbd-client -d /dev/nbd0 on shutdown?

Hmm, looks like everything works just fine as is... So ignore my last request. bnbd is a really interesting alternative to nbd, though. ;)

I would like to use this with bnbd [0]. Can we get an option to set block-size (-block-size), please? [0] https://aur.archlinux.org/packages/bnbd-git/

For completeness, this package should depend on nbd>=3.8 due to the -systemd-mark only being supported from 3.8+

new version uploaded, thanks.

@mus Works for me.

oops, I may have screwed up the parameter order, pushed another fix.

I just pushed this into git, please test it and let me know if it works, I will tag a new version then. I can't test it myself since I'm not using this hook anymore.

The current version of nbd (3.8) now has support for systemd built in. Installing nbd-client as @nbd-client should no longer be necessary. Just add the -m (or -systemd-mark) switch to the nbd-client command in nbd_hook and remove the @ in the nbd_install file and it *should* work. I'll be testing this change soon.

ad encryptssh: https://github.com/nblock/pkgbuild/blob/master/dropbear_initrd_encrypt/encryptssh_hook#L102 The horrible consequences should be obvious. I'm not placing my bets on upstream as of yet as it involves at least three maintainers - mkinitcpio-nbd, encryptssh and nbd. I'll have to iron out the bugs before pushing it. For now, this is a hack I put together and keep using. If it proves to be stable enough (I'm using it on desktops over 10GbE SAN :), I'll try to reach the maintainers. 2.6.32 is the minimum required kernel by archlinux, thankfully. It works flawless in current rolling. The reason being I'm using a desktop system with obscure hardware and redhat-only drivers. The user un-friendliness of this is intentional - whomever will want to use it in different configuration, will have to test it/fix it :)

interesting, thanks for sharing this. The big problem here is that haven't been using this hook myself for a long time because I simply have no need for it anymore. Doesn't mean I won't provide fixes where necessary. ;) After a quick look at you commits, here are my thoughts: - the nbd-client / nbd-driver patches should go upstream. This package is about the hook alone. If upstream doesn't like these patches for whatever reason, I suggest to create separate patched packages for the AUR which this package could (opt-)depend on. Creating AUR packages really isn't that hard actually, have a look at the wiki. ;) And why are you using 2.6.32 LTS? This is a really, really old kernel and not supported by Arch in any way. - tbh, I don't understand all of your commits. e.g. what is "Prevent encryptssh from being stupid" all about? what does this have to do with ssh and the commit just doesn't make any sense to me!? also I don't see why you would need multiple nbd devices in initramfs. one for the root fs should be enough, everything else should be connected to from userspace as soon as root is mounted. but great to see that resuming actually can work, would be great to see this upstream. :)

I've modified your aur to support re-establishing connection, including replay of lost BIOs during network failure/suspend. Also few fixes to make encryptssh less stupid and support multiple nbd devices. Feel free to cherry pick. The juice can be found here: https://github.com/katuma/mkinitcpio-nbd/blob/master/nbd-driver/nbd.patch This has been tested *only* with 2.6.32 LTS, the patch should probably apply even to 3.4. Even suspend to disk worked :) I'm not particularly good at *maintaining* this AUR stuff (i can fix the nbd patch for newer kernels if poked long enough), but it would be nice if you'd eventually rebase to this. I don't plan on making pkgbuilds either.

Thanks for updating this for the mkinitcpio changes!