9p Qemu

naguam

Hello,
I have a NetBSD 10 vm on which I recompiled the kernel to allow virtio-9p.

I'm trying to do some software development for NetBSD, but from my Linux editor and decided to use 9p develop on Linux but from a terminal build and run on NetBSD.

I was able to mount the 9p share with.

mount_9p -cu /dev/vio9p0 /mnt/9p

I had to use the pass-through access mode to be able to write files from the guest, it is not secure, but for pure dev environnement, I accept that.

So I have something working, however I mounted the share inside the guest as root and because 9p does not handle permission, I cannot access the share with a normal user.

I've seen that NetBSD allows users mounts as long as the user have read access, but

mount_9p: puffs_mount: cannot open `/dev/puffs'
mount_9p: puffs_mount: mount_9p: puffs_daemon: Permission denied
Permission denied

I read that user can mount with read access to devices.
How do I give myself access ?
Is chmod in /dev/puffs "OK".
My user is in the wheel group allowing me group access, put /dev/puffs has lrwx------ access rights.

Cheers.

kc9udx

Read this. The article pertains to CDROM filesystems, but the same applies for other devices.

naguam

kc9udx

Thanks for this 🙂 I was able to mount my FS as a user.

Sadly, contrary to what I thought would happen, once mounted the directory still gets root access rights :/

naguam

~~I'm now looking into mount_umap and I've made progress, now the mapping shows is to my user, but for some reason I have now permission denied probably on write permissions, I can only read the files.~~

Ok then it is only about qemu host mode and actual file permissions
But my problem is solved I as able to write files from netbsd.

mount_9p -cu /dev/vio9p0 /mnt/9p
mount_umap -g group.mapfile -u user.mapfile /mnt/9p /home/naguam/9p

naguam

To be fair I'm gonna abandon the idea of using 9p for file sharing.

Tio have user access on either host and guest is too much to ask it seems.
Even if I accept that 9p does not have Unix permissions in the protocol, the virtio method does not work as good as I would have hoped for.

Mapping on QEMU's side is not convenient at all especially when using libvirt.
I'm gonna see if there is a way for me to find another way of sharing host/vm storage.

I did not keep NFS as an option (From NetBSD), because performance was unibelievably slow, and I had problems with remaining files (.nfs000000 and co).

Connecting using SSHFS from host seem to work better at first look, I'm very unsure on the long term as I've seen problems about this on internet.

rvp

naguam

Do what I do: use mc. Assuming your qemu command-line is the standard:

qemu-system-x86_64 -accel nvmm -cpu qemu64 ...
-device virtio-net-pci,netdev=net0 -netdev user,id=net0,hostfwd=tcp::5555-:22

run sshd in the NetBSD VM and connect using mc on Linux:

F9
R
S(h)ell link...
user@localhost:5555

Now the right pane has the VM's files, and the left pane has the host's files.

naguam

rvp I needed access as a "in a filesystem like" manner.

Mounting with SSHFS with FUSE (NetBSD sshd Linux SSHFS client) works well, and I believe that it is what mc does.

I'm running my NetBSD VM on my Fedora Linux laptop using libvirt.

rvp

naguam I needed access as a "in a filesystem like" manner.

I got that--just wanted to point out an alternative. mc is simpler, more convenient, with a lot less drama involved. However, it, like sftp can be slow for multiple files. Here I just use good ol' tar (with judicious --exclude sprinkled):

tar -cf - ... | ssh ... tar -xf - ...

naguam and I believe that it is what mc does.

No, mc rolls its own: either fish or sftp.