http://openvswitch.org/pipermail/dev/2014-July/043119.html
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On 16 July 2014 11:09, Ben Pfaff wrote:
> The dpif provider 'operate' call duplicates all of the features available
> from the 'flow_put', 'flow_del', and 'execute' calls, yielding redundant
> code in providers that support both mechanisms. This change drops the
> latter calls in favor of making
This cleans up the dpif interface to make it more consistent with the
other dpif operations, and allows flows to be fetched in batches.
Signed-off-by: Joe Stringer
---
v2: Require callers to pass down an initialized buffer.
Remove NLM_F_ECHO flag from get request.
Rebase against master
Thanks for the review, I plan to push with the following changes (I noticed
that the other dpif-linux operations - PUT,DEL have redundant memsets, so
deleting them too):
diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c
index e767d9f..c4420f4 100644
--- a/lib/dpif-linux.c
+++ b/lib/dpif-linux.c
@@
Signed-off-by: Joe Stringer
---
datapath/flow_table.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/datapath/flow_table.c b/datapath/flow_table.c
index 7046623..f80e889 100644
--- a/datapath/flow_table.c
+++ b/datapath/flow_table.c
@@ -446,7 +446,7 @@ static void
This allows future patches to reuse the nla error detection code, by
passing their own definitions for expected attribute lengths and types.
Signed-off-by: Joe Stringer
---
datapath/flow_netlink.c | 34 +++---
1 file changed, 19 insertions(+), 15 deletions(-)
diff
Signed-off-by: Joe Stringer
---
datapath/flow_netlink.c | 85 ++-
1 file changed, 39 insertions(+), 46 deletions(-)
diff --git a/datapath/flow_netlink.c b/datapath/flow_netlink.c
index 933b2a1..01f6e49 100644
--- a/datapath/flow_netlink.c
+++ b
On 14 August 2014 10:22, Pravin Shelar wrote:
>
> Can you move this definition next to ovs_key_lens definition.
>
> otherwise looks good.
> Acked-by: Pravin B Shelar
Thanks, I made this change and pushed this to master.
___
dev mailing list
dev@openvs
On 16 August 2014 10:02, Pravin Shelar wrote:
>
> This is good idea. I think we can improve it even further.
> There is nla_parse() function in netlink which does attribute parsing,
> but it is not as strict as OVS needs. Can you look into improving the
> API or adding another API which works for
On 13 August 2014 06:24, Ben Pfaff wrote:
>
> dp_netdev_flow_add() goes to some trouble to log matches as if they
> were provided as datapath flows, even though they aren't. I think
> that match_format() would be simpler and probably less confusing.
In the majority of cases, I agree---however t
Signed-off-by: Joe Stringer
---
lib/dpif.c |1 +
1 file changed, 1 insertion(+)
diff --git a/lib/dpif.c b/lib/dpif.c
index b2a1972..1a32ef6 100644
--- a/lib/dpif.c
+++ b/lib/dpif.c
@@ -840,6 +840,7 @@ dpif_flow_get(struct dpif *dpif,
struct dpif_op *opp;
struct dpif_op op
These tests had the potential to fail due to statistics not updating
before the test script retrieves them. Fix them by waiting until the
next revalidation cycle.
Signed-off-by: Joe Stringer
---
tests/ofproto-dpif.at |6 ++
1 file changed, 6 insertions(+)
diff --git a/tests/ofproto
of the new interface for datapath flow commands. The linux
datapath implementation of these features is yet to come. Previous testing with
a linux datapath implementation using a 64-bit identifier showed reasonable
performance improvements over master. I plan to rework that code and send it
for review next
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 61 ++---
1 file changed, 33 insertions(+), 28 deletions(-)
diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
index 5f1d6d5..bccd7e8 100644
--- a/ofproto/ofproto-dpif
fields optional.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 119 ++---
1 file changed, 51 insertions(+), 68 deletions(-)
diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
index 6f55216..aaf64ed 100644
--- a/ofproto
Signed-off-by: Joe Stringer
---
lib/dpif-netdev.c | 209 ---
lib/flow.h |6 ++
tests/dpif-netdev.at|3 +
tests/ofproto-dpif.at | 20 +++--
tests/ofproto-macros.at |1 +
5 files changed, 164 insertions(+), 75
that two flows hash to the same value, the latter flow will not
be installed.
For backward compatibility, flows that are dumped with no ukey will be hashed
in the same manner at dump time.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 205
Signed-off-by: Joe Stringer
---
lib/dpif-linux.c | 91 +-
1 file changed, 42 insertions(+), 49 deletions(-)
diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c
index f4e78a1..a5f4182 100644
--- a/lib/dpif-linux.c
+++ b/lib/dpif-linux.c
ukey
collections to be read or modified while a revalidator is garbage
collecting it. To protect the ukeys, this patch adds locking on the ukey
collection.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 29 +
1 file changed, 21 insertions(+), 8
Future patches will make use of the 'struct dump_op' in a broader sense,
so this patch renames it to make things a bit clearer.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 95 +
1 file changed, 48 insertions(+), 47 deletion
Add the 128-bit murmurhash by Austin Appleby, for 32-bit systems from:
http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp
Signed-off-by: Joe Stringer
---
There is also a version for 64-bit systems which I haven't tried yet,
mostly because this version provides the same
he handler threads, during flow
setup. This allows future patches to reduce the cost of flow dumping.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 327 -
1 file changed, 222 insertions(+), 105 deletions(-)
diff --git a/ofproto/of
. For each flow dump, the dpif user specifies which fields may be
omitted, allowing the common case to only dump a pair of 128-bit ID and
flow stats.
Signed-off-by: Joe Stringer
---
datapath/linux/compat/include/linux/openvswitch.h | 26 +++
lib/dpctl.c
, which is
raised when flow setup is complete. By checking this flag before
attempting to perform a trylock(), we can reduce lock contention between
handler and revalidator threads.
Signed-off-by: Joe Stringer
---
This patch is particularly relevant after the linux datapath implementation of
UID
maps from the revalidators, and increases
the number of maps used to store ukeys, to reduce contention.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 155 +++--
1 file changed, 87 insertions(+), 68 deletions(-)
diff --git a/ofproto/ofproto
Hi Jarno,
There's been a few iterations. The previous 64-bit version was using a
32-bit hash combined with a thread ID.
The 128-bit hash was based on a suggestion to have sufficient bits to
randomly generate the identifiers, however I've since shifted to using the
hash of the flow key---This allo
On 23 August 2014 10:47, Ben Pfaff wrote:
> MAC learning, like flow entry learning via the "learn" action, should only
> happen if a packet was actually received.
>
> Signed-off-by: Ben Pfaff
Acked-by: Joe Stringer
__
ackets, then no packets could have hit actions for things
such as mac learning/openflow learning. So I'd expect it to be a no-op.
Acked-by: Joe Stringer
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
s entirely from the
> cases that previously only ran if 'may_learn'.
>
> Signed-off-by: Ben Pfaff
>
Acked-by: Joe Stringer
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On 23 August 2014 10:47, Ben Pfaff wrote:
> OpenFlow packet and byte counters have always been something of an
> approximation in Open vSwitch. First, they are updated only periodically.
> Second, they can be misattributed because statistics collection does a
> retranslation and gives the statis
On 23 August 2014 06:29, Ben Pfaff wrote:
> On Thu, Aug 21, 2014 at 05:41:57PM +1200, Joe Stringer wrote:
> > Signed-off-by: Joe Stringer
>
> My reading of the code is that this is redundant because
> dpif_operate() always fills in the flow.
>
Right, this is needed in a l
On 23 August 2014 06:47, Ben Pfaff wrote:
> On Thu, Aug 21, 2014 at 05:41:58PM +1200, Joe Stringer wrote:
> > Signed-off-by: Joe Stringer
>
> As of this patch, there isn't an advantage to using a cmap yet, right?
>
Not yet, no. Cmaps become useful when access is spread
On 23 August 2014 08:01, Ben Pfaff wrote:
> On Thu, Aug 21, 2014 at 05:41:59PM +1200, Joe Stringer wrote:
> > Currently, ukeys are protected during revalidator_sweep__() as only one
> > thread accesses the ukey at a time. This is ensured using barriers:
> > all revalidat
On 23 August 2014 07:59, Ben Pfaff wrote:
> n_umaps is now a hard-coded constant, but a lot of the code assumes
> that it can vary. It would be very slightly simpler to actually use a
> macro or enum to define the fixed size instead of a variable. It
> would also allow better code generation in
On 27 August 2014 06:09, Ben Pfaff wrote:
> On Thu, Aug 21, 2014 at 05:42:02PM +1200, Joe Stringer wrote:
> > Currently, when a revalidator thread first dumps a flow, it creates a
> > 'udpif_key' object and caches a copy of a kernel flow key. This allows
> &g
This function allows netlink attributes to be parsed more strictly,
to check for duplicate attributes, attribute max lengths, and detecting
attributes with only non-zero values.
Signed-off-by: Joe Stringer
---
acinclude.m4 |2 +
datapath/linux/Modules.mk
Signed-off-by: Joe Stringer
---
datapath/flow_netlink.c | 219 +++
1 file changed, 86 insertions(+), 133 deletions(-)
diff --git a/datapath/flow_netlink.c b/datapath/flow_netlink.c
index 69d1919..aa2cece 100644
--- a/datapath/flow_netlink.c
+++ b
I've been having a bit of trouble with ovs after this patch was applied,
ovs-vswitchd locks up on Debian 7.6 (VBox VM). Kernel is 3.2.0-4-amd64 #1
SMP Debian 3.2.60-1+deb7u3 x86_64 GNU/Linux.
I have a fairly typical phy bridge + int bridge setup, with dhclient
running on br-int:
# ovs-vsctl show
RFCv2:
- Revised early patches from v1 feedback
- Rebased
RFCv1:
http://openvswitch.org/pipermail/dev/2014-August/044383.html
Joe Stringer (12):
revalidator: Use 'cmap' for storing ukeys.
revalidator: Protect ukeys with a mutex.
udpif: Separate udpif_key maps from revalidators
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v2: Call ovsrcu_quiesce() unconditionally.
Added ack.
RFC: Initial Post.
---
ofproto/ofproto-dpif-upcall.c | 61 ++---
1 file changed, 33 insertions(+), 28 deletions(-)
diff --git a/ofproto/ofproto
maps from the revalidators, and increases
the number of maps used to store ukeys, to reduce contention.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v2: #define N_UMAPS.
Simplify upcall_unixctl_show() element counting.
Mention that 'ukeys' contains elements of type 's
he handler threads, during flow
setup. This allows future patches to reduce the cost of flow dumping.
Signed-off-by: Joe Stringer
---
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 334 -
1 file changed, 227 insertions(+), 107 deletion
. For each flow dump, the dpif user specifies which fields may be
omitted, allowing the common case to only dump a pair of 128-bit ID and
flow stats.
Signed-off-by: Joe Stringer
---
datapath/linux/compat/include/linux/openvswitch.h | 26 +++
lib/dpctl.c
fields optional.
Signed-off-by: Joe Stringer
---
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 119 ++---
1 file changed, 51 insertions(+), 68 deletions(-)
diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
index
these ukey
collections to be read or modified while a revalidator is garbage
collecting it. To protect the ukeys, this patch adds locking on the ukey
collection.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v2: Added ack.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 29
Future patches will make use of the 'struct dump_op' in a broader sense,
so this patch renames it to make things a bit clearer.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v2: Added ack.
RFC: First post.
---
ofproto/ofproto-dpif-upcal
Add the 128-bit murmurhash by Austin Appleby, for 32-bit systems from:
http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp
Signed-off-by: Joe Stringer
---
There is also a version for 64-bit systems which I haven't tried yet,
mostly because this version provides the same
Signed-off-by: Joe Stringer
---
lib/dpif-netdev.c | 210 ---
lib/flow.h |6 ++
tests/dpif-netdev.at|3 +
tests/ofproto-dpif.at | 20 +++--
tests/ofproto-macros.at |1 +
5 files changed, 165 insertions(+), 75
'handler_install_conflict'.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 198 ++---
ofproto/ofproto-dpif-upcall.h |3 +
ofproto/ofproto-dpif.c| 103 +++--
ofproto/ofproto-dpif.h|2 +
4 files changed, 267
, which is
raised when flow setup is complete. By checking this flag before
attempting to perform a trylock(), we can reduce lock contention between
handler and revalidator threads.
Signed-off-by: Joe Stringer
---
ofproto/ofproto-dpif-upcall.c | 29 +
1 file c
Signed-off-by: Joe Stringer
---
lib/dpif-linux.c | 91 +-
1 file changed, 42 insertions(+), 49 deletions(-)
diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c
index f4e78a1..a5f4182 100644
--- a/lib/dpif-linux.c
+++ b/lib/dpif-linux.c
On 28 August 2014 10:34, Pravin Shelar wrote:
>
>
> This is weird. I could not reproduce this on newer kernel. But I found
> related bug. Can you try attached patch? It may be related.
> I will submit formal patch later.
>
> --Pravin.
>
This patch doesn't fix the issue.
Another slight variatio
On 29 August 2014 15:57, Pravin Shelar wrote:
>
> I have not finished review but I was wonder if you can get rid of the
> attributes bit mask also?
>
Hi Pravin,
I took a look, and I think I can reduce it down to just being inside
match_validate() without much trouble. I don't have a good feel f
An alternative interface, if we want this to be more flexible without
adding new parameters for each possible variation is something like the
following:
diff --git a/datapath/linux/compat/include/linux/netlink.h
b/datapath/linux/compat/include/linux/netlink.h
index f172590..e3d29a2 100644
--- a/da
Add the 128-bit murmurhash by Austin Appleby, for 32-bit systems from:
http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp
Signed-off-by: Joe Stringer
---
v3: Rebase.
There is also a version for 64-bit systems which I haven't tried yet,
mostly because this version provide
. For each flow dump, the dpif user specifies which fields may be
omitted, allowing the common case to only dump a pair of 128-bit ID and
flow stats.
Signed-off-by: Joe Stringer
---
I'm still looking at whether it might be worth trying a 64-bit UID
instead. Any feedback on this would be welcome
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v3: Rebase.
v2: Call ovsrcu_quiesce() unconditionally.
RFC: Initial Post.
---
ofproto/ofproto-dpif-upcall.c | 61 ++---
1 file changed, 33 insertions(+), 28 deletions(-)
diff --git a/ofproto/ofproto-dpif
these ukey
collections to be read or modified while a revalidator is garbage
collecting it. To protect the ukeys, this patch adds locking on the ukey
collection.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v3: Rebase.
v2: No change.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 29
fields optional.
Signed-off-by: Joe Stringer
---
v3: Rebase.
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 118 ++---
1 file changed, 51 insertions(+), 67 deletions(-)
diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif
Future patches will make use of the 'struct dump_op' in a broader sense,
so this patch renames it to make things a bit clearer.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v3: Rebase.
v2: No change.
RFC: First post.
---
ofproto/ofproto-dpif-upcal
'handler_install_conflict'.
Signed-off-by: Joe Stringer
---
v3: Rename the ovs-appctl commands from {en,dis}able-uid->*-terse-dump.
Rebase.
v2: No change.
---
ofproto/ofproto-dpif-upcall.c | 198 ++---
ofproto/ofproto-dpif-upcall.h |3 +
ofproto/o
RFCv3:
- Add datapath implementation
- Minor fixes
- Rebased
RFCv2:
- Revised early patches from v1 feedback
- Add Acks from Ben
- Rebased
RFCv1:
http://openvswitch.org/pipermail/dev/2014-August/044383.html
Joe Stringer (13):
revalidator: Use 'cmap' for storing ukeys.
revalidato
he handler threads, during flow
setup. This allows future patches to reduce the cost of flow dumping.
Signed-off-by: Joe Stringer
---
v3: Rebase.
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 334 -
1 file changed, 227 insertions(+), 107
maps from the revalidators, and increases
the number of maps used to store ukeys, to reduce contention.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v3: Rebase.
v2: #define N_UMAPS.
Simplify upcall_unixctl_show() element counting.
Mention that 'ukeys' contains elemen
, which is
raised when flow setup is complete. By checking this flag before
attempting to perform a trylock(), we can reduce lock contention between
handler and revalidator threads.
Signed-off-by: Joe Stringer
---
v3: No change.
v2: No change.
---
ofproto/ofproto-dpif-upcal
Signed-off-by: Joe Stringer
---
v3: Rebase.
v2: No change.
---
lib/dpif-netdev.c | 211 ---
lib/flow.h |6 ++
tests/dpif-netdev.at|3 +
tests/ofproto-dpif.at | 20 +++--
tests/ofproto-macros.at |1 +
5 files
Signed-off-by: Joe Stringer
---
v3: Rebase.
v2: No change.
---
lib/dpif-linux.c | 91 +-
1 file changed, 42 insertions(+), 49 deletions(-)
diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c
index a32a550..9e4f623 100644
--- a/lib/dpif-linux.c
s returned through the operation. For instance, OVS_UID_F_SKIP_KEY
allows the datapath to skip returning the key.
Signed-off-by: Joe Stringer
---
If the datapath is created using OVS_DP_F_INDEX_BY_UID then a user
starts an older userspace (with no support for UID), the older userspace
will be unable
On 2 September 2014 22:48, Joe Stringer wrote:
> If a datapath is created with the flag OVS_DP_F_INDEX_BY_UID, then an
> additional table_instance is added to the flow_table, which is indexed
> by unique identifiers ("UID"). This can be manipulated using the flow
> key as b
k+0x109/0x1f0
> [] netlink_rcv_skb+0xcb/0xd0
> [] genl_rcv+0x25/0x40
> [] netlink_unicast+0x2f2/0x310
> [] netlink_sendmsg+0x28c/0x3d0
> [] sock_sendmsg+0xef/0x120
> [] ___sys_sendmsg+0x416/0x430
> [] __sys_sendmsg+0x49/0x90
> [] sys_sendmsg+0x19/0x20
> [] system_c
arly patches from v1 feedback
- Add Acks from Ben
- Rebased
RFCv1:
http://openvswitch.org/pipermail/dev/2014-August/044383.html
Joe Stringer (12):
revalidator: Use 'cmap' for storing ukeys.
revalidator: Protect ukeys with a mutex.
udpif: Separate udpif_key maps from revalidators.
upc
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4: No change.
v3: Rebase.
v2: Call ovsrcu_quiesce() unconditionally.
RFC: Initial Post.
---
ofproto/ofproto-dpif-upcall.c | 61 ++---
1 file changed, 33 insertions(+), 28 deletions(-)
diff --git a
maps from the revalidators, and increases
the number of maps used to store ukeys, to reduce contention.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4: Increase N_UMAPS to 512.
v3: Rebase.
v2: #define N_UMAPS.
Simplify upcall_unixctl_show() element counting.
Mention that '
these ukey
collections to be read or modified while a revalidator is garbage
collecting it. To protect the ukeys, this patch adds locking on the ukey
collection.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4: No change.
v3: Rebase.
v2: No change.
RFC: First post.
---
ofproto/ofproto-dpif
Future patches will make use of the 'struct dump_op' in a broader sense,
so this patch renames it to make things a bit clearer.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4: No change.
v3: Rebase.
v2: No change.
RFC: First post.
---
ofproto/ofproto-dpif-upcal
he handler threads, during flow
setup. This allows future patches to reduce the cost of flow dumping.
Signed-off-by: Joe Stringer
---
v4: No change.
v3: Rebase.
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 334 -
1 file changed, 227
, which is
raised when flow setup is complete. By checking this flag before
attempting to perform a trylock(), we can reduce lock contention between
handler and revalidator threads.
Signed-off-by: Joe Stringer
---
v4: Use atomic_read_relaxed().
v3: No change.
v2: No change.
---
ofproto/ofproto-dpi
s returned through the operation. For instance, OVS_UID_F_SKIP_KEY
allows the datapath to skip returning the key.
Signed-off-by: Joe Stringer
---
If the datapath is created using OVS_DP_F_INDEX_BY_UID then a user
starts an older userspace (with no support for UID), the older userspace
will be unable
We previously counted flows that have been installed during the current
dump as duplicates, rather than recognising them as new flows. This
patch separates the counters out for these two cases.
Signed-off-by: Joe Stringer
---
v4: Initial post.
---
ofproto/ofproto-dpif-upcall.c | 13
Add the 128-bit murmurhash by Austin Appleby, for 32-bit systems from:
http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp
Signed-off-by: Joe Stringer
---
v4: Minor comments and style fixes.
v3: Rebase.
---
include/openvswitch/types.h |5 ++
lib/hash.c
simplifies backward
compatibility handling in ofproto-dpif-upcall.
Signed-off-by: Joe Stringer
---
v4: Generate UID in dpif layer and pass up to ofproto-dpif-upcall.
Fix bug where UID-based terse dump wasn't enabled by default.
Skip sending flow key in flow_del.
Fix race condit
fields optional.
Signed-off-by: Joe Stringer
---
v4: No change.
v3: Rebase.
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 118 ++---
1 file changed, 51 insertions(+), 67 deletions(-)
diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto
s are different for each functions). If the hashes are
different, then we have generated the same UID hash for two different
flows, and we should log a warning.
Signed-off-by: Joe Stringer
---
RFC. This is really unlikely, but we don't have any way of detecting
this case currently. This makes
On 9 September 2014 02:10, Jarno Rajahalme wrote:
>
> > On Sep 8, 2014, at 3:23 AM, Joe Stringer wrote:
> >
> > It's highly unlikely that two flows will hash to the same UID. However,
> > we could handle multiple upcalls from the same flow. If we hash the
On 8 September 2014 22:23, Joe Stringer wrote:
> With v4 I've had a fairly good go at testing this series, for performance
> comparison and also compatibility with mixed older/newer
> datapath/userspace.
> Master appears to be able to revalidate roughly 130-140K flows per
enl_lock+0x17/0x20
> [] genl_rcv+0x16/0x40
> [] netlink_unicast+0x2f2/0x310
> [] netlink_ack+0x109/0x1f0
> [] netlink_rcv_skb+0xcb/0xd0
> [] genl_rcv+0x25/0x40
> [] netlink_unicast+0x2f2/0x310
> [] netlink_sendmsg+0x28c/0x3d0
> [] sock_sendmsg+0xef/0x120
> [] ___sy
This function allows netlink attributes to be parsed more strictly,
to check for additional constraints beyond those that nla_parse() checks
for. Passing flags=0 to nla_parse_strict() implies the same behaviour as
nla_parse().
Signed-off-by: Joe Stringer
---
v2: Change nla_parse_strict
Reduce duplicate code by using nla_policy and nla_parse_strict().
Signed-off-by: Joe Stringer
---
v2: Remove attrs bitmasks from most functions.
Remove key/mask wrappers for parse_nlattrs().
Rebase against flag-based nla_parse_strict().
---
datapath/flow_netlink.c | 420
improvement in the supported number of flows when operating around
the flow_limit.
Signed-off-by: Joe Stringer
---
ofproto/ofproto.h |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/ofproto/ofproto.h b/ofproto/ofproto.h
index 1b8709a..d60b198 100644
--- a/ofproto/ofproto.h
+++ b
Thanks, applied to master.
On 15 September 2014 09:41, Ethan Jackson wrote:
> Acked-by: Ethan Jackson
>
>
> On Thu, Sep 11, 2014 at 3:03 PM, Joe Stringer
> wrote:
> > The datapath max_idle value determines how long to wait before deleting
> > an idle datapath fl
fixes
- Rebased
RFCv2:
- Revised early patches from v1 feedback
- Add Acks from Ben
- Rebased
RFCv1:
http://openvswitch.org/pipermail/dev/2014-August/044383.html
Joe Stringer (12):
revalidator: Use 'cmap' for storing ukeys.
revalidator: Protect ukeys with a mutex.
udpif: Separate udpi
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4-v5: No change.
v3: Rebase.
v2: Call ovsrcu_quiesce() unconditionally.
RFC: Initial Post.
---
ofproto/ofproto-dpif-upcall.c | 61 ++---
1 file changed, 33 insertions(+), 28 deletions(-)
diff --git a
these ukey
collections to be read or modified while a revalidator is garbage
collecting it. To protect the ukeys, this patch adds locking on the ukey
collection.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4-v5: No change.
v3: Rebase.
v2: No change.
RFC: First post.
---
ofproto/ofproto
maps from the revalidators, and increases
the number of maps used to store ukeys, to reduce contention.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v5: No change.
v4: Increase N_UMAPS to 512.
v3: Rebase.
v2: #define N_UMAPS.
Simplify upcall_unixctl_show() element counting.
Mention
Add the 128-bit murmurhash by Austin Appleby, for 32-bit systems from:
http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp
Signed-off-by: Joe Stringer
---
v5: No change.
v4: Minor comments and style fixes.
v3: Rebase.
---
include/openvswitch/types.h |5 ++
lib/hash.c
Future patches will make use of the 'struct dump_op' in a broader sense,
so this patch renames it to make things a bit clearer.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v4-v5: No change.
v3: Rebase.
v2: No change.
RFC: First post.
---
ofproto/ofproto-dpif-upcal
he handler threads, during flow
setup. This allows future patches to reduce the cost of flow dumping.
Signed-off-by: Joe Stringer
---
v4-v5: No change.
v3: Rebase.
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 334 -
1 file changed, 227
fields optional.
Signed-off-by: Joe Stringer
---
v5: Switch back to checking that the mask is more specific.
v4: No change.
v3: Rebase.
v2: Rebase.
RFC: First post.
---
ofproto/ofproto-dpif-upcall.c | 105 -
1 file changed, 52 insertions(+), 53 deletions
simplifies backward
compatibility handling in ofproto-dpif-upcall.
Signed-off-by: Joe Stringer
---
v5: Always pass flow_key down to dpif, to improve error logging.
Fix extraneous netflow_unref.
Fix testsuite failure.
Rebase.
v4: Generate UID in dpif layer and pass up to ofproto-d
s returned through the operation. For instance, OVS_UID_F_SKIP_KEY
allows the datapath to skip returning the key.
Signed-off-by: Joe Stringer
---
v5: No change.
v4: Fix memory leaks.
Log when triggering the older userspace issue above.
v3: Initial post.
---
datapath/
1 - 100 of 2028 matches
Mail list logo
