[ovs-dev] Bug#630325: Bug#630325: Bug#630325: openvswitch: FTBFS on big-endian architectures

[Ben Pfaff]

On Tue, Jun 14, 2011 at 10:05:35PM -0700, Ben Pfaff wrote:
> On Wed, Jun 15, 2011 at 02:04:39PM +0900, Simon Horman wrote:
> > On Tue, Jun 14, 2011 at 07:56:52PM -0700, Ben Pfaff wrote:
> > > On Wed, Jun 15, 2011 at 10:51:24AM +0900, Simon Horman wrote:
> > > > Unfortunately I'm not familiar enough with the output to determine
> > > > the problem without digging much further. Perhaps you have some ideas.
> > > > 
> > > > The detailed output from sparc is below.
> > > > I believe that the output is similar if not the same
> > > > on the other three architectures.
> > > 
> > > Yes, I looked at this today.  I'll send out a fix for it tomorrow.
> > > 
> > > It's a bug in the testsuite, I think, not in OVS itself.
> > 
> > Understood.
> > 
> > Could you CC me or this bug on the fix? Then I can push it into
> > an upload to Debian.Org and the buildds can verify the change.
> 
> Of course.  Will do.

Simon, looking closer, this was fixed on branch-1.1 by the following
commit, made on April 18.  (The fix is to sort the output of
dump-flows.)  This fix is already in OVS 1.1.1.  This bug is reported
against OVS 1.1.0, so it should be obsoleted by the OVS version
currently in Debian.  Indeed, when I look at the current buildd status
at https://buildd.debian.org/status/package.php?p=openvswitch&suite=sid
I don't see any relevant failures.

I think we can close this.

Thanks,

Ben.

--8<--------------------------cut here-------------------------->8--

From: Ben Pfaff <blp at nicira.com>
Date: Mon, 18 Apr 2011 10:11:43 -0700
Subject: [PATCH] ofp-util: Properly handle "tun_id"s in tun_id_from_cookie
 flows.

Just setting the tun_id field isn't enough--it's also necessary to set
the tun_id_mask.  Otherwise the call to cls_rule_zero_wildcarded_fields()
at the end of ofputil_cls_rule_from_match() will zero out the tun_id again.

This was broken by commit 8368c090cab "Implement arbitrary bitwise masks
for tun_id field" back in January.  (This makes me wonder whether we can
drop support for tun_id_from_cookie now.)

Reported-by: Dan Wendlandt <dan at nicira.com>
---
 lib/ofp-util.c   |    2 +-
 tests/ofproto.at |    8 ++++++--
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/lib/ofp-util.c b/lib/ofp-util.c
index cc448bc..5c95270 100644
--- a/lib/ofp-util.c
+++ b/lib/ofp-util.c
@@ -136,7 +136,7 @@ ofputil_cls_rule_from_match(const struct ofp_match *match,
     wc->nw_dst_mask = ofputil_wcbits_to_netmask(ofpfw >> OFPFW_NW_DST_SHIFT);
 
     if (flow_format == NXFF_TUN_ID_FROM_COOKIE && !(ofpfw & NXFW_TUN_ID)) {
-        rule->flow.tun_id = htonll(ntohll(cookie) >> 32);
+        cls_rule_set_tun_id(rule, htonll(ntohll(cookie) >> 32));
     }
 
     if (ofpfw & OFPFW_DL_DST) {
diff --git a/tests/ofproto.at b/tests/ofproto.at
index 9506756..fc7ff57 100644
--- a/tests/ofproto.at
+++ b/tests/ofproto.at
@@ -48,10 +48,14 @@ AT_CHECK([ovs-ofctl dump-flows br0 | STRIP_XIDS], [0], [NXST_FLOW reply:
 ])
 AT_CHECK([echo 'in_port=1,actions=0' | ovs-ofctl add-flows br0 -])
 AT_CHECK([ovs-ofctl add-flow br0 in_port=0,actions=1])
-AT_CHECK([ovs-ofctl dump-flows br0 | STRIP_XIDS | STRIP_DURATION], [0], [dnl
-NXST_FLOW reply:
+dnl Tests for a bug in which ofproto ignored tun_id in tun_id_from_cookie
+dnl flow_mod commands.
+AT_CHECK([ovs-ofctl add-flow -F tun_id_from_cookie br0 tun_id=1,actions=mod_vlan_vid:4])
+AT_CHECK([ovs-ofctl dump-flows br0 | STRIP_XIDS | STRIP_DURATION | sort], [0], [dnl
  cookie=0x0, duration=?s, table_id=0, n_packets=0, n_bytes=0, in_port=1 actions=output:0
  cookie=0x0, duration=?s, table_id=0, n_packets=0, n_bytes=0, in_port=65534 actions=output:1
+ cookie=0x100000000, duration=?s, table_id=0, n_packets=0, n_bytes=0, tun_id=0x1 actions=mod_vlan_vid:4
+NXST_FLOW reply:
 ])
 AT_CHECK([ovs-ofctl del-flows br0])
 AT_CHECK([ovs-ofctl dump-flows br0 | STRIP_XIDS], [0], [NXST_FLOW reply:
-- 
1.7.4.4