cmap_find_protected calculated wrong h2 hash which causing entries with
duplicated id inserted into the cmap.
Reported-at:
https://mail.openvswitch.org/pipermail/ovs-discuss/2018-December/047945.html
Signed-off-by: Zang MingJie
---
lib/cmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion
at 4:25 PM Zang MingJie wrote:
> ---
> lib/cmap.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/lib/cmap.c b/lib/cmap.c
> index cb9cd32ab..c9eef3f4a 100644
> --- a/lib/cmap.c
> +++ b/lib/cmap.c
> @@ -568,7 +568,7 @@ cmap_find_protected(const
---
lib/cmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/cmap.c b/lib/cmap.c
index cb9cd32ab..c9eef3f4a 100644
--- a/lib/cmap.c
+++ b/lib/cmap.c
@@ -568,7 +568,7 @@ cmap_find_protected(const struct cmap *cmap, uint32_t hash)
{
struct cmap_impl *impl =
Hi, we found a serious problem where one pmd is stop working, I want to
share the problem and find solution here.
vswitchd log:
2018-09-13T23:36:44.377Z|40269235|dpif_netdev(pmd45)|WARN|Packet dropped.
Max recirculation depth exceeded.
On Thu, Sep 20, 2018 at 8:47 AM Darrell Ball wrote:
>
>
> On Fri, Sep 14, 2018 at 1:46 AM, Zang MingJie
> wrote:
>
>> >> > Did you notice this check ?
>> >> >
>> >> > if (src->state < CT_DPIF_TCPS_SYN_SENT) {
>> >
<- syn+ack <-invalid
malformed packet will set wrong seq_lo and seq_hi to the state, preventing
following packets pass ct.
>
>>
>>
>> >
>> >
>> >
>> >>
>> >>
>> >> Signed-off-by: Zang MingJie
>> >&
On Thu, Sep 13, 2018 at 2:55 AM Darrell Ball wrote:
>
> Thanks for looking MingJie
>
>
> On Wed, Sep 12, 2018 at 2:16 AM, Zang MingJie wrote:
>>
>> When encounter an invalid packet, all changes made by the packet should
>> be reverted. Currently an invalid packe
When encounter an invalid packet, all changes made by the packet should
be reverted. Currently an invalid packet can change the seq number while
the connection is in SEQ_RECV state.
Signed-off-by: Zang MingJie
---
lib/conntrack-tcp.c | 10 --
1 file changed, 8 insertions(+), 2 deletions
Hi:
Currently with user space datapath, only internal tap device can be
used to communicate with kernel. Because packets must be copied from
user space to kernel, the throughput of tap implementation is very
slow, not suitable for data plane. Now DPDK provides some alternative
solutions.
1. KNI
I don't think the patch will resolve the problem. Once ctb->lock is
released, other thread may have chance to acquire the lock and modify
ctb. In general, ctb->lock can not be released in this function,
another approach is needed.
On Wed, Aug 29, 2018 at 3:31 PM Darrell Ball wrote:
>
> nat_clean
While developing application using ovs userspace conntrack, we found
some bugs worth mention here.
1. conntrack_clean may causes ovs crash.
conntrack_clean function iterators through all buckets, and free
entries in the bucket with bucket lock, but when releasing a NAT
connection, inside
Hi:
I found that in struct ofputil_meter_band, rate and burst are all 32 bits,
the max burst is:
for bandwidth: 4GB = 32Gb
for pps: 4G packets/1000 = 4M packets
In our high performance setup, 4M packets burst is easily achievable, so I
suggest change these variable to uint64_t
Confirmed, can be easily reproduced using described method.
Using ovs 2.6.2
On 01/23/2017 11:58 PM, Vidyasagara Guntaka via dev wrote:
Hi Ben,
We could reproduce this with the latest version 2.6.1. When we compiled the
code, we removed -O2 from CFLAGS. This seems to make it happen more
Yes, It is better to use the bit-field. I'll update the patch after
the refereed patch has been applied
On Tue, Apr 4, 2017 at 10:03 PM, Kevin Traynor <ktray...@redhat.com> wrote:
> On 03/15/2017 08:46 AM, Zang MingJie wrote:
>> dpdk-strip-vlan option specifies whether strip vl
dpdk-strip-vlan option specifies whether strip vlan for the dpdk interface.
Signed-off-by: Zang MingJie <zealot0...@gmail.com>
---
lib/netdev-dpdk.c| 23 ++-
vswitchd/vswitch.xml | 7 +++
2 files changed, 29 insertions(+), 1 deletion(-)
diff --git a/lib/
The push/pop header of flow based tunnel is generated dynamically by flows, so
there won't be any tunnel header associated to netdev.
The patch fixes that tnl_port_cache doesn't contain any flow based tunnel.
Signed-off-by: Zang MingJie <zealot0...@gmail.com>
---
lib/dpif-netdev.c | 2 +
16 matches
Mail list logo
