Re: [Qemu-devel] [Consult] nfs-vsocks support

[Stefan Hajnoczi <stefanha@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 3362 bytes --]

On Wed, Mar 28, 2018 at 07:25:21PM +0800, Liu, Jing2 wrote:
> On 3/26/2018 6:12 PM, Liu, Jing2 wrote:
> > Hi Stefan,
> > 
> > Thank you very much for the response! It truly gave me much help.
> > 
> > On 3/24/2018 12:02 AM, Stefan Hajnoczi wrote:
> [...]
> > > > Could you give me some help and thanks in advance!
> > > 
> > > Here is a script that launches nfsd and runs a guest:
> > > https://github.com/stefanha/linux/commit/38cbc15661a6dd44b69c4f318091f2047d707035#diff-05f3fe8941076453942a8c059b409009
> > > 
> 
> Using the script, I met some problems making the
> initramfs.gz. It confused me a lot... I don't have busybox under
> /usr/bin or /usr/sbin. Where to get it?

It may be easier to use a full guest disk image instead of the disk
image generation in my go.sh script.  The initramfs is very specific to
the development machine I was using at the time.  The main thing to look
at the in the go.sh script is how the NFS services are launched on the
host.

Busybox is a package that is available on most Linux distributions.

Before you spend more time on this, NFS over AF_VSOCK is only suitable
for developers right now.  It is not ready for production and I am not
supporting it for end users.  So if you want to hack on NFS, please go
ahead, but otherwise it's probably not what you want.

Stefan

> (host)$./go.sh nfs_tcp
> 
> File /usr/sbin/busybox could not be opened for reading
>  line 55
> File /lib64/libtirpc.so.3 could not be opened for reading
>  line 306
> File nc-vsock could not be opened for reading
>  line 321
> File ../netperf-2.7.0/src/netserver could not be opened for reading
>  line 325
> 
> It results in guest kernel_panic when gdb launches the guest:
> 
> [    0.425738] Failed to execute /init (error -2)
> [    0.426074] Kernel panic - not syncing: No working init found.  Try
> passing .
> [    0.427077] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.13.0+ #3
> [    0.427507] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
> rel-4
> [    0.428354] Call Trace:
> [    0.428538]  dump_stack+0x63/0x89
> [    0.428774]  ? rest_init+0x60/0xc0
> [    0.429017]  panic+0xeb/0x245
> [    0.429225]  ? putname+0x53/0x60
> [    0.429455]  ? rest_init+0xc0/0xc0
> [    0.429722]  kernel_init+0xf1/0x104
> [    0.430017]  ret_from_fork+0x25/0x30
> 
> Looking forward to the kind reply.
> BTW, for the following questions that I sent two days ago, I have some
> ideas in mind and it could be ignored now.
> 
> Jing
> 
> > I have updated both host and guest kernel with vsock-nfsd repo and
> > installed nfs-utils in both host and guest successfully.
> > 
> > There're two different manuals in
> > https://www.spinics.net/lists/linux-nfs/msg64563.html(Quickstart 3~7)
> > and nfs-utils (README step 3 DAEMON STARTUP ORDER).
> > Which one should be the best usage?
> > 
> > For the first quickstart manual step 7-Mount the export from the guest,
> > two questions as follows.
> > 1. why hypervisor's CID is 2? I didn't notice when it is specified?
> > 2. Though I did step 3~7 successfully, I can't see the synchronistical
> > changing on host folder /export/ and guest /mnt? So how to test the
> > communcation between host and guest on the folder?
> > 
> > Thanks again!
> > 
> > Jing
> [...]
> > > 
> > > Stefan
> > > 
> > 
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]