Page MenuHomeFreeBSD

Add netmap support to bhyve
ClosedPublic

Authored by gnn on Jan 8 2016, 4:54 AM.

Details

Summary

This patch originally submitted by btw@. Updated to work on HEAD. Provides basic netmap support in the guest.

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Automatic diff as part of commit; lint not applicable.
Unit
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

gnn retitled this revision from to Add netmap support to bhyve.
gnn updated this object.
gnn edited the test plan for this revision. (Show Details)
gnn added reviewers: grehan, btw.

I tried to test this patch on -HEAD, but it caused kernel panic. It seems that there is a bug in the kqueue support code of netmap. And I can reproduce the kernel panic with this simple program on -HEAD:

#include <stdio.h>
#include <assert.h>

#include <sys/types.h>
#include <sys/event.h>
#include <sys/time.h>

#ifndef NETMAP_WITH_LIBS
#define NETMAP_WITH_LIBS
#endif
#include <net/netmap_user.h>

static int
kq_add(int kq, int fd)
{
	struct kevent changes[1];
	int ret;

	EV_SET(&changes[0], fd, EVFILT_READ, EV_ADD, 0, 0, NULL);
	ret = kevent(kq, changes, 1, NULL, 0, NULL);
	assert(ret != -1);

	printf("[%s] success\n", __func__);

	return (ret);
}

static void
kq_wait(int kq)
{
	struct kevent events[1];
	int ret;

	ret = kevent(kq, NULL, 0, events, 1, NULL);
	assert(ret != -1);

	printf("[%s] success\n", __func__);
}

int main(void)
{
	const char *ifname = "vale0:vm1";
	struct nm_desc *d;
	int kq;

	d = nm_open(ifname, NULL, 0, 0);
	assert(d != NULL);

	kq = kqueue();
	assert(kq != -1);

	kq_add(kq, d->fd);
	kq_wait(kq);

	return (0);
}

And below is part of the crash summary:

freebsd dumped core - see /var/crash/vmcore.0

Fri Jan  8 22:19:39 CST 2016

FreeBSD freebsd 11.0-CURRENT FreeBSD 11.0-CURRENT #2 d1f6105(master)-dirty: Fri Jan  8 21:55:33 CST 2016     btw@freebsd:/usr/obj/root/freebsd/sys/GENERIC  amd64

......

panic: mutex nm_kn_lock not owned at /root/freebsd/sys/kern/kern_event.c:2073
cpuid = 1
KDB: stack backtrace:
db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe011744c420
vpanic() at vpanic+0x182/frame 0xfffffe011744c4a0
panic() at panic+0x43/frame 0xfffffe011744c500
__mtx_assert() at __mtx_assert+0xbf/frame 0xfffffe011744c510
knlist_add() at knlist_add+0x20/frame 0xfffffe011744c540
netmap_kqfilter() at netmap_kqfilter+0x101/frame 0xfffffe011744c580
devfs_kqfilter_f() at devfs_kqfilter_f+0x81/frame 0xfffffe011744c5d0
kqueue_register() at kqueue_register+0x5bf/frame 0xfffffe011744c670
kqueue_kevent() at kqueue_kevent+0xc8/frame 0xfffffe011744c840
kern_kevent_fp() at kern_kevent_fp+0x99/frame 0xfffffe011744c890
kern_kevent() at kern_kevent+0x9f/frame 0xfffffe011744c8f0
sys_kevent() at sys_kevent+0x11c/frame 0xfffffe011744c9a0
amd64_syscall() at amd64_syscall+0x2db/frame 0xfffffe011744cab0
Xfast_syscall() at Xfast_syscall+0xfb/frame 0xfffffe011744cab0
--- syscall (363, FreeBSD ELF64, sys_kevent), rip = 0x80095bc7a, rsp = 0x7fffffffe988, rbp = 0x7fffffffe9e0 ---
KDB: enter: panic

......

I don't see that here either in the VM or outside of one. Here is the uname of my VM:

FreeBSD odevbox 11.0-CURRENT FreeBSD 11.0-CURRENT #0 2c6aacd(dtrace-gnn)-dirty: Wed Dec 16 04:55:40 EST 2015 root@odevbox:/usr/obj/mnt/minion/exported/github/CADETS/freebsd/sys/GENERIC-NODEBUG amd64

Are you seeing this just on your VM or on generic netmap on HEAD?

I'm testing on the master branch of the repo cloned from https://github.com/freebsd/freebsd.git with GENERIC conf on a desktop computer. And I can also reproduce this panic in VirtualBox with the image downloaded from http://mirrors.aliyun.com/freebsd/snapshots/VM-IMAGES/11.0-CURRENT/amd64/20151217/FreeBSD-11.0-CURRENT-amd64-20151217-r292413.vhd.xz

If you're seeing this on an image without the patch then it's not the patch that's the problem. If that's the case then please open a PR against netmap and, if you like assign it to me but it ought to not block committing this work.

I have verified the bug you mention is unrelated to this patch. It happens on vanilla HEAD. I'd like to commit this patch and then look at the bug you mention.

Yeah, the bug is unrelated to this patch. I'll open a PR against netmap and won't let it block committing work! And thank you very much for being willing to commit this patch! ^_^

I will try to revert netmap and test this patch!

This revision was automatically updated to reflect the committed changes.