qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] VirtIO issues after 83d768b5640946b7da55ce8335509df297e


From: Brian Rak
Subject: Re: [Qemu-devel] VirtIO issues after 83d768b5640946b7da55ce8335509df297e2c7cd
Date: Wed, 15 Feb 2017 10:17:49 -0500
User-agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:45.0) Gecko/20100101 Thunderbird/45.7.1

On 2/10/2017 3:35 AM, Brad Campbell wrote:
On 03/02/17 01:14, Brian Rak wrote:
We recently upgraded to qemu 2.8.0, and noticed a bunch of issues with
various non-linux operating systems.

I was able to bisect this down to
83d768b5640946b7da55ce8335509df297e2c7cd being the commit that breaks
things (0687c37c5eeef8580b31cc6e1202d874833ae38a which is immediately
prior still works ok).

The issue here seems to be that guests stop receiving packets, and this
only happens with vhost_net enabled.


I've been consistently able to reproduce this with by booting the
OpenBSD 5.8 ISO (
http://mirrors.evowise.com/pub/OpenBSD/5.8/amd64/install58.iso ), and
pressing 'A' at the first prompt.  It'll attempt to do a DHCP lease, and
will succeed prior to 83d7.  After that commit, it'll continue to
transit DHCP requests, but never actually receive the reply.

So far, we've seen this same problem manifest on the following operating
systems:

* OpenBSD 5.8
* pfSense 2.2.6 (seems to be FreeBSD 10.1 underneath)
* Windows XP
* "9front" (fork of Plan 9)

More modern operating systems did work, including

* Various Linux distributions (CentOS 6+, Ubuntu 12.04+)
* Modern windows distributions (Server 2012)
* OpenBSD 6

I was pointed to this thread -
https://lists.gnu.org/archive/html/qemu-devel/2017-01/msg04532.html ,
and if I apply that patch network connectivity begins working normally
again. (I am running under x86-64 btw)



I raised this on the 18th of Jan too. https://lists.gnu.org/archive/html/qemu-devel/2017-01/msg03467.html

I found that it would fail with any linux guest compiled without CONFIG_GENERIC_MSI_IRQ.

Thanks for the pointer to the patch. I can apply that and set all the XP guests back to virtio networking. (Had to change them to ne2k_pci to get networking).

Regards,
Brad

I'm not 100% sure if the patch is production-ready or not.. there was some discussion about it, but it's a bit over my head. We had applied it just to confirm it fixed the issue we were seeing, but it never left our development environment. We just ended up rolling back qemu, breaking all sorts of legacy guests wasn't really something we were willing to do.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]