Due to the race condition in userspace, there is chance that two
overlapping megaflows could be installed in datapath. And this
causes userspace unable to delete the less inclusive megaflow flow
even after it timeout, since the flow_del logic will stop at the
first match of masked flow.
This commit fixes the bug by making the kernel flow_del and flow_get
logic check all masks in that case.
Introduced by 03f0d916a (openvswitch: Mega flow implementation).
Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Incorrect struct name was confusing, even though otherwise
inconsequental.
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Masks are inserted when flows are inserted to the table, so it is
logical to correspondingly remove masks when flows are removed from
the table, in ovs_flow_table_remove().
This allows ovs_flow_free() to be called without locking, which will
be used by later patches.
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Keep kernel flow stats for each NUMA node rather than each (logical)
CPU. This avoids using the per-CPU allocator and removes most of the
kernel-side OVS locking overhead otherwise on the top of perf reports
and allows OVS to scale better with higher number of threads.
With 9 handlers and 4 revalidators netperf TCP_CRR test flow setup
rate doubles on a server with two hyper-threaded physical CPUs (16
logical cores each) compared to the current OVS master. Tested with
non-trivial flow table with a TCP port match rule forcing all new
connections with unique port numbers to OVS userspace. The IP
addresses are still wildcarded, so the kernel flows are not considered
as exact match 5-tuple flows. This type of flows can be expected to
appear in large numbers as the result of more effective wildcarding
made possible by improvements in OVS userspace flow classifier.
Perf results for this test (master):
Events: 305K cycles
+ 8.43% ovs-vswitchd [kernel.kallsyms] [k] mutex_spin_on_owner
+ 5.64% ovs-vswitchd [kernel.kallsyms] [k] __ticket_spin_lock
+ 4.75% ovs-vswitchd ovs-vswitchd [.] find_match_wc
+ 3.32% ovs-vswitchd libpthread-2.15.so [.] pthread_mutex_lock
+ 2.61% ovs-vswitchd [kernel.kallsyms] [k] pcpu_alloc_area
+ 2.19% ovs-vswitchd ovs-vswitchd [.] flow_hash_in_minimask_range
+ 2.03% swapper [kernel.kallsyms] [k] intel_idle
+ 1.84% ovs-vswitchd libpthread-2.15.so [.] pthread_mutex_unlock
+ 1.64% ovs-vswitchd ovs-vswitchd [.] classifier_lookup
+ 1.58% ovs-vswitchd libc-2.15.so [.] 0x7f4e6
+ 1.07% ovs-vswitchd [kernel.kallsyms] [k] memset
+ 1.03% netperf [kernel.kallsyms] [k] __ticket_spin_lock
+ 0.92% swapper [kernel.kallsyms] [k] __ticket_spin_lock
...
And after this patch:
Events: 356K cycles
+ 6.85% ovs-vswitchd ovs-vswitchd [.] find_match_wc
+ 4.63% ovs-vswitchd libpthread-2.15.so [.] pthread_mutex_lock
+ 3.06% ovs-vswitchd [kernel.kallsyms] [k] __ticket_spin_lock
+ 2.81% ovs-vswitchd ovs-vswitchd [.] flow_hash_in_minimask_range
+ 2.51% ovs-vswitchd libpthread-2.15.so [.] pthread_mutex_unlock
+ 2.27% ovs-vswitchd ovs-vswitchd [.] classifier_lookup
+ 1.84% ovs-vswitchd libc-2.15.so [.] 0x15d30f
+ 1.74% ovs-vswitchd [kernel.kallsyms] [k] mutex_spin_on_owner
+ 1.47% swapper [kernel.kallsyms] [k] intel_idle
+ 1.34% ovs-vswitchd ovs-vswitchd [.] flow_hash_in_minimask
+ 1.33% ovs-vswitchd ovs-vswitchd [.] rule_actions_unref
+ 1.16% ovs-vswitchd ovs-vswitchd [.] hindex_node_with_hash
+ 1.16% ovs-vswitchd ovs-vswitchd [.] do_xlate_actions
+ 1.09% ovs-vswitchd ovs-vswitchd [.] ofproto_rule_ref
+ 1.01% netperf [kernel.kallsyms] [k] __ticket_spin_lock
...
There is a small increase in kernel spinlock overhead due to the same
spinlock being shared between multiple cores of the same physical CPU,
but that is barely visible in the netperf TCP_CRR test performance
(maybe ~1% performance drop, hard to tell exactly due to variance in
the test results), when testing for kernel module throughput (with no
userspace activity, handful of kernel flows).
On flow setup, a single stats instance is allocated (for the NUMA node
0). As CPUs from multiple NUMA nodes start updating stats, new
NUMA-node specific stats instances are allocated. This allocation on
the packet processing code path is made to never block or look for
emergency memory pools, minimizing the allocation latency. If the
allocation fails, the existing preallocated stats instance is used.
Also, if only CPUs from one NUMA-node are updating the preallocated
stats instance, no additional stats instances are allocated. This
eliminates the need to pre-allocate stats instances that will not be
used, also relieving the stats reader from the burden of reading stats
that are never used.
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
The 5-tuple optimization becomes unnecessary with a later per-NUMA
node stats patch. Remove it first to make the changes easier to
grasp.
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
In few functions, const formal parameters are assigned or cast to
non-const.
These changes suppress warnings if compiled with -Wcast-qual.
Signed-off-by: Daniele Di Proietto <daniele.di.proietto@gmail.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
ovs_flow_free() is not called under ovs-lock during packet
execute path (ovs_packet_cmd_execute()). Since packet execute
does not touch flow->mask, there is no need to take that
lock either. So move assert in case where flow->mask is checked.
Found by code inspection.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Both mega flow mask's reference counter and per flow table mask list
should only be accessed when holding ovs_mutex() lock. However
this is not true with ovs_flow_table_flush(). The patch fixes this bug.
Reported-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
memory allocated by kmem_cache_alloc() should be freed using
kmem_cache_free(), not kfree().
Fixes: e298e50570 ('openvswitch: Per cpu flow stats.')
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jesse Gross says:
====================
[GIT net-next] Open vSwitch
Open vSwitch changes for net-next/3.14. Highlights are:
* Performance improvements in the mechanism to get packets to userspace
using memory mapped netlink and skb zero copy where appropriate.
* Per-cpu flow stats in situations where flows are likely to be shared
across CPUs. Standard flow stats are used in other situations to save
memory and allocation time.
* A handful of code cleanups and rationalization.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
As we're only doing a kfree() anyway in the RCU callback, we can
simply use kfree_rcu, which does the same job, and remove the
function rcu_free_sw_flow_mask_cb() and rcu_free_acts_callback().
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
With mega flow implementation ovs flow can be shared between
multiple CPUs which makes stats updates highly contended
operation. This patch uses per-CPU stats in cases where a flow
is likely to be shared (if there is a wildcard in the 5-tuple
and therefore likely to be spread by RSS). In other situations,
it uses the current strategy, saving memory and allocation time.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Flow lookup can happen either in packet processing context or userspace
context but it was annotated as requiring RCU read lock to be held. This
also allows OVS mutex to be held without causing warnings.
Reported-by: Justin Pettit <jpettit@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Reviewed-by: Thomas Graf <tgraf@redhat.com>
API changes only for code readability. No functional chnages.
This patch removes the underscored version. Added a new API
ovs_flow_tbl_lookup_stats() that returns the n_mask_hits.
Reported by: Ben Pfaff <blp@nicira.com>
Reviewed-by: Thomas Graf <tgraf@redhat.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Currently OVS uses jhash2() for calculating flow hashes in its
internal flow_hash() function. The performance of the flow_hash()
function is critical, as the input data can be hundreds of bytes
long.
OVS is largely deployed in x86_64 based datacenters. Therefore,
we argue that the performance critical fast path of OVS should
exploit underlying CPU features in order to reduce the per packet
processing costs. We replace jhash2 with the hash implementation
provided by the kernel hash lib, which exploits the crc32l
instruction to achieve high performance
Our patch greatly reduces the hash footprint from ~200 cycles of
jhash2() to around ~90 cycles in case of ovs_flow_hash_crc()
(measured with rdtsc over maximum length flow keys on an i7 Intel
CPU).
Additionally, we wrote a microbenchmark to stress the flow table
performance. The benchmark inserts random flows into the flow
hash and then performs lookups. Our hash deployed on a CRC32
capable CPU reduces the lookup for 1000 flows, 100 masks from
~10,100us to ~6,700us, for example.
Thus, simply use the newly introduced arch_fast_hash2() as a
drop-in replacement.
Signed-off-by: Francesco Fusco <ffusco@redhat.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Thomas Graf <tgraf@redhat.com>
Acked-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Flow->hash can be used to detect hash collisions and avoid flow key
compare in flow lookup.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Collect mega flow mask stats. ovs-dpctl show command can be used to
display them for debugging and performance tuning.
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Hides mega-flow implementation in flow_table.c rather than
datapath.c.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
ovs-flow rehash does not touch mega flow list. Following patch
moves it dp struct datapath. Avoid one extra indirection for
accessing mega-flow list head on every packet receive.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Over the time datapath.c and flow.c has became pretty large files.
Following patch restructures functionality of component into three
different components:
flow.c: contains flow extract.
flow_netlink.c: netlink flow api.
flow_table.c: flow table api.
This patch restructures code without changing logic.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
