LinuxLists.cc - [PATCH] net/sock: move memory_allocated over to percpu

2018-09-06 19:20:34

Subject: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

Today these are all global shared variables per protocol, and in
particular tcp_memory_allocated can get hot on a system with
large number of CPUs and a substantial number of connections.

Moving it over to a per-cpu variable makes it significantly cheaper,
and the added overhead when summing up the percpu copies is still smaller
than the cost of having a hot cacheline bouncing around.

Signed-off-by: Olof Johansson <[email protected]>
---
crypto/af_alg.c | 10 ++++++++--
include/net/sctp/sctp.h | 3 ++-
include/net/sock.h | 12 ++++++------
include/net/tcp.h | 2 +-
include/net/udp.h | 2 +-
net/core/sock.c | 5 ++++-
net/decnet/af_decnet.c | 3 ++-
net/ipv4/tcp.c | 3 ++-
net/ipv4/udp.c | 4 +++-
net/sctp/protocol.c | 6 ++++++
net/sctp/socket.c | 2 +-
11 files changed, 36 insertions(+), 16 deletions(-)

diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index b053179e0bc5..1fd75a709d7b 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -29,7 +29,7 @@ struct alg_type_list {
struct list_head list;
};

-static atomic_long_t alg_memory_allocated;
+static struct percpu_counter alg_memory_allocated;

static struct proto alg_proto = {
.name = "ALG",
@@ -1183,13 +1183,19 @@ static int __init af_alg_init(void)
if (err)
goto out;

- err = sock_register(&alg_family);
+ err = percpu_counter_init(&alg_memory_allocated, 0, GFP_KERNEL);
if (err != 0)
goto out_unregister_proto;

+ err = sock_register(&alg_family);
+ if (err != 0)
+ goto out_free_percpu;
+
out:
return err;

+out_free_percpu:
+ percpu_counter_destroy(&alg_memory_allocated);
out_unregister_proto:
proto_unregister(&alg_proto);
goto out;
diff --git a/include/net/sctp/sctp.h b/include/net/sctp/sctp.h
index 8c2caa370e0f..270579cf310b 100644
--- a/include/net/sctp/sctp.h
+++ b/include/net/sctp/sctp.h
@@ -36,7 +36,7 @@
* Sridhar Samudrala <[email protected]>
* Ardelle Fan <[email protected]>
* Ryan Layer <[email protected]>
- * Kevin Gao <[email protected]>
+ * Kevin Gao <[email protected]>
*/

#ifndef __net_sctp_h__
@@ -114,6 +114,7 @@ __poll_t sctp_poll(struct file *file, struct socket *sock,
void sctp_sock_rfree(struct sk_buff *skb);
void sctp_copy_sock(struct sock *newsk, struct sock *sk,
struct sctp_association *asoc);
+extern struct percpu_counter sctp_memory_allocated;
extern struct percpu_counter sctp_sockets_allocated;
int sctp_asconf_mgmt(struct sctp_sock *, struct sctp_sockaddr_entry *);
struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
diff --git a/include/net/sock.h b/include/net/sock.h
index 433f45fc2d68..45aed5e84b5d 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1114,7 +1114,7 @@ struct proto {
/* Memory pressure */
void (*enter_memory_pressure)(struct sock *sk);
void (*leave_memory_pressure)(struct sock *sk);
- atomic_long_t *memory_allocated; /* Current allocated memory. */
+ struct percpu_counter *memory_allocated; /* Current allocated memory. */
struct percpu_counter *sockets_allocated; /* Current number of sockets. */
/*
* Pressure flag: try to collapse.
@@ -1237,19 +1237,19 @@ static inline bool sk_under_memory_pressure(const struct sock *sk)
static inline long
sk_memory_allocated(const struct sock *sk)
{
- return atomic_long_read(sk->sk_prot->memory_allocated);
+ return percpu_counter_sum_positive(sk->sk_prot->memory_allocated);
}

-static inline long
+static inline void
sk_memory_allocated_add(struct sock *sk, int amt)
{
- return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
+ percpu_counter_add(sk->sk_prot->memory_allocated, amt);
}

static inline void
sk_memory_allocated_sub(struct sock *sk, int amt)
{
- atomic_long_sub(amt, sk->sk_prot->memory_allocated);
+ percpu_counter_sub(sk->sk_prot->memory_allocated, amt);
}

static inline void sk_sockets_allocated_dec(struct sock *sk)
@@ -1277,7 +1277,7 @@ proto_sockets_allocated_sum_positive(struct proto *prot)
static inline long
proto_memory_allocated(struct proto *prot)
{
- return atomic_long_read(prot->memory_allocated);
+ return percpu_counter_sum_positive(prot->memory_allocated);
}

static inline bool
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 770917d0caa7..2df1754cf3ab 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -248,7 +248,7 @@ extern long sysctl_tcp_mem[3];
#define TCP_RACK_STATIC_REO_WND 0x2 /* Use static RACK reo wnd */
#define TCP_RACK_NO_DUPTHRESH 0x4 /* Do not use DUPACK threshold in RACK */

-extern atomic_long_t tcp_memory_allocated;
+extern struct percpu_counter tcp_memory_allocated;
extern struct percpu_counter tcp_sockets_allocated;
extern unsigned long tcp_memory_pressure;

diff --git a/include/net/udp.h b/include/net/udp.h
index 8482a990b0bb..9e0d9f7091a0 100644
--- a/include/net/udp.h
+++ b/include/net/udp.h
@@ -97,7 +97,7 @@ static inline struct udp_hslot *udp_hashslot2(struct udp_table *table,

extern struct proto udp_prot;

-extern atomic_long_t udp_memory_allocated;
+extern struct percpu_counter udp_memory_allocated;

/* sysctl variables for udp */
extern long sysctl_udp_mem[3];
diff --git a/net/core/sock.c b/net/core/sock.c
index 3730eb855095..0a755f6c8942 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -2394,9 +2394,12 @@ EXPORT_SYMBOL(sk_wait_data);
int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
{
struct proto *prot = sk->sk_prot;
- long allocated = sk_memory_allocated_add(sk, amt);
+ long allocated;
bool charged = true;

+ sk_memory_allocated_add(sk, amt);
+ allocated = sk_memory_allocated(sk);
+
if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
!(charged = mem_cgroup_charge_skmem(sk->sk_memcg, amt)))
goto suppress_allocation;
diff --git a/net/decnet/af_decnet.c b/net/decnet/af_decnet.c
index 7d6ff983ba2c..f88af9ae4474 100644
--- a/net/decnet/af_decnet.c
+++ b/net/decnet/af_decnet.c
@@ -156,7 +156,7 @@ static const struct proto_ops dn_proto_ops;
static DEFINE_RWLOCK(dn_hash_lock);
static struct hlist_head dn_sk_hash[DN_SK_HASH_SIZE];
static struct hlist_head dn_wild_sk;
-static atomic_long_t decnet_memory_allocated;
+static struct percpu_counter decnet_memory_allocated;

static int __dn_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen, int flags);
static int __dn_getsockopt(struct socket *sock, int level, int optname, char __user *optval, int __user *optlen, int flags);
@@ -2356,6 +2356,7 @@ static int __init decnet_init(void)
int rc;

printk(banner);
+ percpu_counter_init(&decnet_memory_allocated, 0, GFP_KERNEL);

rc = proto_register(&dn_proto, 1);
if (rc != 0)
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 8c4235c098fd..eb6531ba6bd3 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -289,7 +289,7 @@ EXPORT_SYMBOL_GPL(tcp_orphan_count);
long sysctl_tcp_mem[3] __read_mostly;
EXPORT_SYMBOL(sysctl_tcp_mem);

-atomic_long_t tcp_memory_allocated; /* Current allocated memory. */
+struct percpu_counter tcp_memory_allocated; /* Current allocated memory. */
EXPORT_SYMBOL(tcp_memory_allocated);

#if IS_ENABLED(CONFIG_SMC)
@@ -3834,6 +3834,7 @@ void __init tcp_init(void)
BUILD_BUG_ON(sizeof(struct tcp_skb_cb) >
FIELD_SIZEOF(struct sk_buff, cb));

+ percpu_counter_init(&tcp_memory_allocated, 0, GFP_KERNEL);
percpu_counter_init(&tcp_sockets_allocated, 0, GFP_KERNEL);
percpu_counter_init(&tcp_orphan_count, 0, GFP_KERNEL);
inet_hashinfo_init(&tcp_hashinfo);
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index f4e35b2ff8b8..6ec5d2f68ae7 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -122,7 +122,7 @@ EXPORT_SYMBOL(udp_table);
long sysctl_udp_mem[3] __read_mostly;
EXPORT_SYMBOL(sysctl_udp_mem);

-atomic_long_t udp_memory_allocated;
+struct percpu_counter udp_memory_allocated;
EXPORT_SYMBOL(udp_memory_allocated);

#define MAX_UDP_PORTS 65536
@@ -2923,6 +2923,8 @@ void __init udp_init(void)

__udp_sysctl_init(&init_net);

+ percpu_counter_init(&udp_memory_allocated, 0, GFP_KERNEL);
+
/* 16 spinlocks per cpu */
udp_busylocks_log = ilog2(nr_cpu_ids) + 4;
udp_busylocks = kmalloc(sizeof(spinlock_t) << udp_busylocks_log,
diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c
index e948db29ab53..ca59ca0dc740 100644
--- a/net/sctp/protocol.c
+++ b/net/sctp/protocol.c
@@ -1391,6 +1391,10 @@ static __init int sctp_init(void)
if (!sctp_chunk_cachep)
goto err_chunk_cachep;

+ status = percpu_counter_init(&sctp_memory_allocated, 0, GFP_KERNEL);
+ if (status)
+ goto err_percpu_memory_init;
+
status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
if (status)
goto err_percpu_counter_init;
@@ -1559,6 +1563,8 @@ static __init int sctp_init(void)
err_ehash_alloc:
percpu_counter_destroy(&sctp_sockets_allocated);
err_percpu_counter_init:
+ percpu_counter_destroy(&sctp_memory_allocated);
+err_percpu_memory_init:
kmem_cache_destroy(sctp_chunk_cachep);
err_chunk_cachep:
kmem_cache_destroy(sctp_bucket_cachep);
diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index f73e9d38d5ba..60d55573baa5 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -107,7 +107,7 @@ static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
enum sctp_socket_type type);

static unsigned long sctp_memory_pressure;
-static atomic_long_t sctp_memory_allocated;
+struct percpu_counter sctp_memory_allocated;
struct percpu_counter sctp_sockets_allocated;

static void sctp_enter_memory_pressure(struct sock *sk)
--
2.11.0

2018-09-06 19:33:58

by Eric Dumazet

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

On Thu, Sep 6, 2018 at 12:21 PM Olof Johansson <[email protected]> wrote:
>
> Today these are all global shared variables per protocol, and in
> particular tcp_memory_allocated can get hot on a system with
> large number of CPUs and a substantial number of connections.
>
> Moving it over to a per-cpu variable makes it significantly cheaper,
> and the added overhead when summing up the percpu copies is still smaller
> than the cost of having a hot cacheline bouncing around.

I am curious. We never noticed contention on this variable, at least for TCP.

Please share some numbers with us.

2018-09-07 03:32:57

by Herbert Xu

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

On Thu, Sep 06, 2018 at 12:33:58PM -0700, Eric Dumazet wrote:
> On Thu, Sep 6, 2018 at 12:21 PM Olof Johansson <[email protected]> wrote:
> >
> > Today these are all global shared variables per protocol, and in
> > particular tcp_memory_allocated can get hot on a system with
> > large number of CPUs and a substantial number of connections.
> >
> > Moving it over to a per-cpu variable makes it significantly cheaper,
> > and the added overhead when summing up the percpu copies is still smaller
> > than the cost of having a hot cacheline bouncing around.
>
> I am curious. We never noticed contention on this variable, at least for TCP.

Yes these variables are heavily amortised so I'm surprised that
they would cause much contention.

> Please share some numbers with us.

Indeed.

Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

2018-09-07 06:20:15

by Olof Johansson

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

Hi,

On Thu, Sep 6, 2018 at 8:32 PM, Herbert Xu <[email protected]> wrote:
> On Thu, Sep 06, 2018 at 12:33:58PM -0700, Eric Dumazet wrote:
>> On Thu, Sep 6, 2018 at 12:21 PM Olof Johansson <[email protected]> wrote:
>> >
>> > Today these are all global shared variables per protocol, and in
>> > particular tcp_memory_allocated can get hot on a system with
>> > large number of CPUs and a substantial number of connections.
>> >
>> > Moving it over to a per-cpu variable makes it significantly cheaper,
>> > and the added overhead when summing up the percpu copies is still smaller
>> > than the cost of having a hot cacheline bouncing around.
>>
>> I am curious. We never noticed contention on this variable, at least for TCP.
>
> Yes these variables are heavily amortised so I'm surprised that
> they would cause much contention.
>
>> Please share some numbers with us.
>
> Indeed.

Certainly, just had to collect them again.

This is on a dual xeon box, with ~150-200k TCP connections. I see
about .7% CPU spent in __sk_mem_{reduce,raise}_allocated in the
inlined atomic ops, most of those in reduce.

Call path for reduce is practically all from tcp_write_timer on softirq:

__sk_mem_reduce_allocated
tcp_write_timer
call_timer_fn
run_timer_softirq
__do_softirq
irq_exit
smp_apic_timer_interrupt
apic_timer_interrupt
cpuidle_enter_state

With this patch, I see about .18+.11+.07 = .36% in percpu-related
functions called from the same __sk_mem functions.

Now, that's a halving of cycles samples on that specific setup. The
real difference though, is on another platform where atomics are more
expensive. There, this makes a significant difference. Unfortunately,
I can't share specifics but I think this change stands on its own on
the dual xeon setup as well, maybe with slightly less strong wording
on just how hot the variable/line happens to be.

-Olof

2018-09-07 07:03:13

by Eric Dumazet

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

On Thu, Sep 6, 2018 at 11:20 PM Olof Johansson <[email protected]> wrote:
>
> Hi,
>
> On Thu, Sep 6, 2018 at 8:32 PM, Herbert Xu <[email protected]> wrote:
> > On Thu, Sep 06, 2018 at 12:33:58PM -0700, Eric Dumazet wrote:
> >> On Thu, Sep 6, 2018 at 12:21 PM Olof Johansson <[email protected]> wrote:
> >> >
> >> > Today these are all global shared variables per protocol, and in
> >> > particular tcp_memory_allocated can get hot on a system with
> >> > large number of CPUs and a substantial number of connections.
> >> >
> >> > Moving it over to a per-cpu variable makes it significantly cheaper,
> >> > and the added overhead when summing up the percpu copies is still smaller
> >> > than the cost of having a hot cacheline bouncing around.
> >>
> >> I am curious. We never noticed contention on this variable, at least for TCP.
> >
> > Yes these variables are heavily amortised so I'm surprised that
> > they would cause much contention.
> >
> >> Please share some numbers with us.
> >
> > Indeed.
>
> Certainly, just had to collect them again.
>
> This is on a dual xeon box, with ~150-200k TCP connections. I see
> about .7% CPU spent in __sk_mem_{reduce,raise}_allocated in the
> inlined atomic ops, most of those in reduce.
>
> Call path for reduce is practically all from tcp_write_timer on softirq:
>
> __sk_mem_reduce_allocated
> tcp_write_timer
> call_timer_fn
> run_timer_softirq
> __do_softirq
> irq_exit
> smp_apic_timer_interrupt
> apic_timer_interrupt
> cpuidle_enter_state
>
> With this patch, I see about .18+.11+.07 = .36% in percpu-related
> functions called from the same __sk_mem functions.
>
> Now, that's a halving of cycles samples on that specific setup. The
> real difference though, is on another platform where atomics are more
> expensive. There, this makes a significant difference. Unfortunately,
> I can't share specifics but I think this change stands on its own on
> the dual xeon setup as well, maybe with slightly less strong wording
> on just how hot the variable/line happens to be.

Problem is : we have platforms with more than 100 cpus, and
sk_memory_allocated() cost will be too expensive,
especially if the host is under memory pressure, since all cpus will
touch their private counter.

per cpu variables do not really scale, they were ok 10 years ago when
no more than 16 cpus were the norm.

I would prefer change TCP to not aggressively call
__sk_mem_reduce_allocated() from tcp_write_timer()

Ideally only tcp_retransmit_timer() should attempt to reduce forward
allocations, after recurring timeout.

Note that after 20c64d5cd5a2bdcdc8982a06cb05e5e1bd851a3d ("net: avoid
sk_forward_alloc overflows")
we have better control over sockets having huge forward allocations.

Something like :

diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 7fdf222a0bdfe9775970082f6b5dcdcc82b2ae1a..7e2e17cde9b6a9be835edfac26b64f4ce9411538
100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -505,6 +505,8 @@ void tcp_retransmit_timer(struct sock *sk)
mib_idx = LINUX_MIB_TCPTIMEOUTS;
}
__NET_INC_STATS(sock_net(sk), mib_idx);
+ } else {
+ sk_mem_reclaim(sk);
}

tcp_enter_loss(sk);
@@ -576,11 +578,11 @@ void tcp_write_timer_handler(struct sock *sk)

if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN)) ||
!icsk->icsk_pending)
- goto out;
+ return;

if (time_after(icsk->icsk_timeout, jiffies)) {
sk_reset_timer(sk, &icsk->icsk_retransmit_timer,
icsk->icsk_timeout);
- goto out;
+ return;
}

tcp_mstamp_refresh(tcp_sk(sk));
@@ -602,9 +604,6 @@ void tcp_write_timer_handler(struct sock *sk)
tcp_probe_timer(sk);
break;
}
-
-out:
- sk_mem_reclaim(sk);
}

2018-09-07 07:21:46

by Eric Dumazet

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

On Fri, Sep 7, 2018 at 12:03 AM Eric Dumazet <[email protected]> wrote:

> Problem is : we have platforms with more than 100 cpus, and
> sk_memory_allocated() cost will be too expensive,
> especially if the host is under memory pressure, since all cpus will
> touch their private counter.
>
> per cpu variables do not really scale, they were ok 10 years ago when
> no more than 16 cpus were the norm.
>
> I would prefer change TCP to not aggressively call
> __sk_mem_reduce_allocated() from tcp_write_timer()
>
> Ideally only tcp_retransmit_timer() should attempt to reduce forward
> allocations, after recurring timeout.
>
> Note that after 20c64d5cd5a2bdcdc8982a06cb05e5e1bd851a3d ("net: avoid
> sk_forward_alloc overflows")
> we have better control over sockets having huge forward allocations.
>
> Something like :

Or something less risky :

diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 7fdf222a0bdfe9775970082f6b5dcdcc82b2ae1a..0aee80b6966cb2898e46350c761f9eb431ff1206
100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -604,7 +604,8 @@ void tcp_write_timer_handler(struct sock *sk)
}

out:
- sk_mem_reclaim(sk);
+ if (tcp_under_memory_pressure(sk))
+ sk_mem_reclaim(sk);
}

static void tcp_write_timer(struct timer_list *t)

2018-09-08 17:02:42

by Olof Johansson

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

Hi,

On Fri, Sep 7, 2018 at 12:21 AM, Eric Dumazet <[email protected]> wrote:
> On Fri, Sep 7, 2018 at 12:03 AM Eric Dumazet <[email protected]> wrote:
>
>> Problem is : we have platforms with more than 100 cpus, and
>> sk_memory_allocated() cost will be too expensive,
>> especially if the host is under memory pressure, since all cpus will
>> touch their private counter.
>>
>> per cpu variables do not really scale, they were ok 10 years ago when
>> no more than 16 cpus were the norm.
>>
>> I would prefer change TCP to not aggressively call
>> __sk_mem_reduce_allocated() from tcp_write_timer()
>>
>> Ideally only tcp_retransmit_timer() should attempt to reduce forward
>> allocations, after recurring timeout.
>>
>> Note that after 20c64d5cd5a2bdcdc8982a06cb05e5e1bd851a3d ("net: avoid
>> sk_forward_alloc overflows")
>> we have better control over sockets having huge forward allocations.
>>
>> Something like :
>
> Or something less risky :

I gave both of these patches a run, and neither do as well on the
system that has slower atomics. :(

The percpu version:

8.05% workload [kernel.vmlinux]
[k] __do_softirq
7.04% swapper [kernel.vmlinux]
[k] cpuidle_enter_state
5.54% workload [kernel.vmlinux]
[k] _raw_spin_unlock_irqrestore
1.66% swapper [kernel.vmlinux]
[k] __do_softirq
1.55% workload [kernel.vmlinux]
[k] finish_task_switch
1.24% swapper [kernel.vmlinux]
[k] finish_task_switch
1.07% workload [kernel.vmlinux]
[k] net_rx_action

The first patch from you still has significant amount of time spent in
the atomics paths (non-inlined versions used):

7.87% workload [kernel.vmlinux]
[k] __ll_sc_atomic64_sub
7.48% workload [kernel.vmlinux]
[k] __do_softirq
5.05% workload [kernel.vmlinux]
[k] _raw_spin_unlock_irqrestore
2.42% workload [kernel.vmlinux]
[k] __ll_sc_atomic64_add_return
1.49% swapper [kernel.vmlinux]
[k] cpuidle_enter_state
1.31% workload [kernel.vmlinux]
[k] finish_task_switch
1.09% workload [kernel.vmlinux]
[k] tcp_sendmsg_locked
1.08% workload [kernel.vmlinux]
[k] __arch_copy_from_user
1.02% workload [kernel.vmlinux]
[k] net_rx_action

I think a lot of the overhead from percpu approach can be alleviated
if we can use percpu_counter_read() instead of _sum() (i.e. no need to
iterate through the local per-cpu recent delta). I don't know the TCP
stack well enough to tell where it's OK to use a bit of slack in the
numbers though -- by default count will at most be off by 32*online
cpus. Might not be a significant number in reality.

-Olof

2018-09-09 18:38:37

by Eric Dumazet

[permalink] [raw]

Subject: Re: [PATCH] net/sock: move memory_allocated over to percpu_counter variables

On Sat, Sep 8, 2018 at 10:02 AM Olof Johansson <[email protected]> wrote:
>
> Hi,
>
> On Fri, Sep 7, 2018 at 12:21 AM, Eric Dumazet <[email protected]> wrote:
> > On Fri, Sep 7, 2018 at 12:03 AM Eric Dumazet <[email protected]> wrote:
> >
> >> Problem is : we have platforms with more than 100 cpus, and
> >> sk_memory_allocated() cost will be too expensive,
> >> especially if the host is under memory pressure, since all cpus will
> >> touch their private counter.
> >>
> >> per cpu variables do not really scale, they were ok 10 years ago when
> >> no more than 16 cpus were the norm.
> >>
> >> I would prefer change TCP to not aggressively call
> >> __sk_mem_reduce_allocated() from tcp_write_timer()
> >>
> >> Ideally only tcp_retransmit_timer() should attempt to reduce forward
> >> allocations, after recurring timeout.
> >>
> >> Note that after 20c64d5cd5a2bdcdc8982a06cb05e5e1bd851a3d ("net: avoid
> >> sk_forward_alloc overflows")
> >> we have better control over sockets having huge forward allocations.
> >>
> >> Something like :
> >
> > Or something less risky :
>
> I gave both of these patches a run, and neither do as well on the
> system that has slower atomics. :(
>
> The percpu version:
>
> 8.05% workload [kernel.vmlinux]
> [k] __do_softirq
> 7.04% swapper [kernel.vmlinux]
> [k] cpuidle_enter_state
> 5.54% workload [kernel.vmlinux]
> [k] _raw_spin_unlock_irqrestore
> 1.66% swapper [kernel.vmlinux]
> [k] __do_softirq
> 1.55% workload [kernel.vmlinux]
> [k] finish_task_switch
> 1.24% swapper [kernel.vmlinux]
> [k] finish_task_switch
> 1.07% workload [kernel.vmlinux]
> [k] net_rx_action
>
> The first patch from you still has significant amount of time spent in
> the atomics paths (non-inlined versions used):
>
> 7.87% workload [kernel.vmlinux]
> [k] __ll_sc_atomic64_sub

The second patch I gave should not enter this path at all, please try it.

> 7.48% workload [kernel.vmlinux]
> [k] __do_softirq
> 5.05% workload [kernel.vmlinux]
> [k] _raw_spin_unlock_irqrestore
> 2.42% workload [kernel.vmlinux]
> [k] __ll_sc_atomic64_add_return
> 1.49% swapper [kernel.vmlinux]
> [k] cpuidle_enter_state
> 1.31% workload [kernel.vmlinux]
> [k] finish_task_switch
> 1.09% workload [kernel.vmlinux]
> [k] tcp_sendmsg_locked
> 1.08% workload [kernel.vmlinux]
> [k] __arch_copy_from_user
> 1.02% workload [kernel.vmlinux]
> [k] net_rx_action
>
> I think a lot of the overhead from percpu approach can be alleviated
> if we can use percpu_counter_read() instead of _sum() (i.e. no need to
> iterate through the local per-cpu recent delta). I don't know the TCP
> stack well enough to tell where it's OK to use a bit of slack in the
> numbers though -- by default count will at most be off by 32*online
> cpus. Might not be a significant number in reality.

2018-09-18 09:37:22

by Chen, Rong A

[permalink] [raw]

Subject: [LKP] [net/sock] b99259a614: netperf.Throughput_Mbps -6.6% regression

Greeting,

FYI, we

commit: url:
in testcase: netperf
on test with following parameters:

ip: ipv4
runtime: 300s
nr_threads: 200%
cluster: cs-localhost
send_size: 5K
test: TCP_SENDFILE
ucode: 0x25
cpufreq_governor: performance

test-description: test-url: In addition
+---------------- | testcase: | test machine | test parameters | | | | | +---------------- | testcase: | test machine | test parameters | | | | | +----------------

Details are as below:
-----------------

To reproduce:

git clone bin/lkp run job.yaml

================= cluster/compiler/ cs-localhost/gc
commit:
6da410d97f b99259a614
6da410d97ffa486e ---------------- fail:runs | 1:4 1:4 :4 1:4 1:4 1:4 1:4 0:4 %stddev \ 7990 127841 1874683 643.25 1792 144.93 1099277 9.363e+08 30722 25888 ± 31% 25888 ± 31% 7151 ± 4% 7409 ± 3% 38618 ± 10% 0.13 ± 15% 161917 ± 4% 3215895 ± 15% 38639 ± 10% 1.305e+12 3.00 3.92e+10 0.69 ± 9% 2.1e+10 9364894 1.31 0.09 2.02e+09 2.201e+12 1.38e+12 4.827e+08 ± 8% 6.749e+12 0.76 7207 18.44 18.46 18.25 18.17 13.45 12.68 11.95 9.53 ± 2% 9.35 ± 2% 9.50 10.73 ± 2% 10.56 ± 2% 2.99 ± 3% 4.94 ± 2% 0.70 ± 14% 3.04 ± 2% 2.62 ± 3% 2.31 ± 2% 1.52 ± 5% 1.14 ± 4% 1.88 ± 2% 1.11 ± 16% 3.21 0.74 ± 23% 1.19 ± 7% 1.20 ± 3% 1.19 ± 5% 0.80 ± 4% 0.90 ± 6% 0.71 ± 2% 1.07 ± 3% 0.74 ± 6% 0.63 ± 3% 0.70 ± 3% 0.78 ± 3% 0.82 ± 3% 5.58 5.35 0.91 ± 4% 90.58 0.00 0.00 0.00 0.00 0.00 0.00 0.00 70.71 68.26 0.00 57.51 54.10 36.84 35.89 35.00 30.95 29.68 28.59 28.87 27.07 21.40 20.54 0.00 1.22 ± 4% 1.32 ± 3% 18.44 18.46 18.25 18.19 15.06 13.47 12.00 9.58 ± 2% 10.09 9.42 ± 2% 9.68 11.60 ± 2% 11.47 ± 2% 6.05 ± 2% 3.54 ± 3% 2.50 ± 3% 3.05 ± 2% 2.06 2.70 ± 2% 3.39 ± 3% 2.10 ± 2% 1.29 ± 13% 2.57 ± 4% 2.37 2.34 ± 2% 1.21 ± 5% 3.67 2.72 ± 4% 1.61 ± 4% 0.86 ± 20% 0.47 ± 18% 0.99 ± 5% 0.41 ± 21% 0.52 ± 16% 0.56 ± 15% 1.49 ± 2% 0.32 ± 25% 1.77 ± 4% 1.10 ± 3% 0.70 ± 11% 0.96 ± 2% 0.88 ± 2% 0.85 ± 2% 0.46 ± 4% 0.37 ± 5% 0.44 ± 3% 0.12 ± 27% 0.20 ± 16% 0.16 ± 13% 0.21 ± 7% 0.37 ± 3% 0.25 ± 7% 0.06 ± 11% 0.06 ± 13% 0.09 ± 18% 0.09 ± 11% 0.04 ± 60% 0.91 ± 4% 0.11 ± 7% 0.33 ± 10% 3.19 ± 2% 0.55 ± 12% 3.07 ± 2% 0.18 ± 27% 90.71 0.08 ± 13% 0.00 0.00 0.00 0.00 70.92 68.42 57.60 0.25 ± 13% 54.16 36.88 35.98 35.13 31.00 29.75 28.64 28.94 27.21 21.48 20.69 0.00 1.34 ± 3% 1.46 ± 3% 1.21 ± 3% 9.27 3.41 ± 3% 3.12 ± 3% 2.01 ± 2% 2.52 2.42 3.66 1.18 ± 5% 1.62 ± 3% 1.19 ± 2% 1.44 ± 5% 0.89 ± 5% 1.12 ± 2% 1.02 ± 3% 0.81 ± 7% 0.71 ± 7% 0.89 ± 6% 0.14 ± 9% 0.75 ± 4% 0.97 ± 3% 0.82 0.29 ± 8% 0.42 ± 5% 1.57 0.12 ± 27% 0.15 ± 7% 0.15 ± 14% 0.17 ± 9% 0.32 0.14 ± 10% 0.23 ± 9% 0.22 ± 9% 0.11 ± 19% 0.11 ± 4% 0.11 ± 7% 0.01 ±173% 0.09 ± 7% 0.15 ± 16% 0.00 0.00 0.00 0.24 ± 14% 0.00 0.00

8100 +-+------- | 8000 +-+++.++. + + 7900 +-+ ++ | 7800 +-+ | 7700 +-+ | 7600 +-+ 7500 +-+ | 7400 +-+ O O OO OO O 7300 +-+------- 130000 | 128000 +-+++.++. : +.+ |+ ++ 126000 +-+ | 124000 +-+ | 122000 +-+ | 120000 +-+ | 118000 O-+OO O |O O 116000 9.5e+08 9.4e+08 +-+ |+.++.++ : + 9.3e+08 +-+ +.+ 9.2e+08 +-+ | 9.1e+08 +-+ 9e+08 +-+ 8.9e+08 +-+ | 8.8e+08 +-+ 8.7e+08 +-+ O O OO O 8.6e+08 +O+ O 8.5e+08 152 +-+-------- 150 +-+ | 148 +-+ 146 +-+ +. + 144 +-++ + + 142 +-+ | + 140 +-+ 138 +-+ 136 +-+ O 134 +-+ | O 132 OO+OO OO 130 +-+-------- 1840 +-+-O----- |O O O 1830 O-+ O O | | 1820 +-+ | 1810 +-+ | 1800 +-+ | +.+ + | .++.+ 1790 +-+ | 1780 +-+------- 1.12e+06 | 1.1e+06 +-+++. : 1.08e+06 +-+ | 1.06e+06 +-+ | 1.04e+06 +-+ | 1.02e+06 +-+ 1e+06 +-+ |O O O 980000 O-+OO O | 960000 [*] bisect-good sample
[O] bisect-bad sample

***************** lkp-bdw-ep2: ================= cluster/compiler/ cs-localhost/gc
commit:
6da410d97f b99259a614
6da410d97ffa486e ---------------- fail:runs | :4 1:4 11:4 4:4 31:4 1:4 %stddev \ 3140615 ± 2% 71377 ± 2% 6877 ± 4% 2795 7688 727.56 ± 3% 9.335e+08 ± 3% 9.422e+08 ± 2% 497.45 143.48 25025 ± 50% 823677 14.68 37.84 4.73 ± 3% 231917 ± 7% 458545 ± 3% 467255 ± 4% 53.00 12208228 ± 3% 197402 9.422e+08 ± 2% 2090351 ± 22% 6824307 8055584 7.34e+09 1.833e+09 ± 3% 3.942e+08 ± 63% 558983 ± 49% 1.824e+08 ± 3% 34485351 ± 5% 1997 71.97 1.833e+09 ± 3% 27.42 ± 2% 556509 ± 50% 1.46 ± 64% 27.21 ± 2% 270.14 15.33 360260 ± 11% 2840994 922773 ± 6% 360260 ± 11% 2840994 922773 ± 6% 2804769 ± 2% 2905503 2731218 ± 3% 29631 ± 28% 0.58 ± 4% 10.43 ± 9% 29606 ± 28% 387.56 356998 ± 25% 58785 ± 15% 0.46 ± 15% 2.25 ± 8% 4.17 ± 14% 21059143 ± 3% 21578752 ± 2% 20097461 ± 6% 262381 ± 31% 5.784e+12 ± 2% 2.76 1.599e+11 ± 2% 0.73 ± 8% 2.527e+09 ± 7% 3.482e+11 3.733e+09 ± 3% 1.93 5.633e+13 7.772e+09 ± 4% 8.521e+12 ± 2% 0.00 ± 8% 1.095e+08 ± 9% 5.49e+12 ± 2% 10.06 ± 5% 9.169e+09 ± 5% 8.203e+10 ± 3% 2.915e+13 ± 2% 0.52 801554 7.603e+08 ± 5% 93.23 ± 3% 1.072e+09 ± 5% 80515351 ± 49% 801557 30931 15.38 15.18 13.40 13.14 8.62 ± 6% 10.50 8.98 ± 2% 5.03 ± 12% 9.64 9.42 4.54 ± 13% 44.22 4.46 ± 13% 43.77 43.76 7.84 ± 2% 43.66 8.92 7.58 ± 2% 7.46 ± 2% 8.60 8.49 6.10 5.95 4.93 4.64 3.65 4.00 3.24 ± 2% 3.27 2.52 2.48 ± 2% 2.23 ± 4% 2.24 2.16 ± 2% 1.94 ± 2% 1.66 ± 5% 1.57 ± 5% 1.55 ± 2% 1.65 ± 2% 1.52 1.57 ± 2% 1.46 1.29 ± 5% 0.89 ± 2% 1.26 ± 5% 1.46 1.20 ± 5% 1.34 1.20 1.00 ± 2% 1.06 ± 2% 1.10 ± 3% 0.95 ± 3% 30.84 29.71 49.06 48.63 32.67 32.48 31.55 30.98 20.10 19.82 19.32 17.83 15.21 15.07 14.83 14.73 14.31 13.95 13.17 12.85 12.43 12.27 11.92 11.37 10.29 0.00 0.00 0.00 0.00 15.40 15.20 13.61 13.40 13.15 8.64 ± 6% 10.53 8.98 ± 2% 5.04 ± 12% 9.65 9.43 4.54 ± 13% 44.22 44.22 44.20 4.46 ± 13% 43.77 8.93 7.59 ± 2% 8.61 8.51 6.16 4.93 4.64 3.70 4.09 3.32 ± 2% 3.28 2.98 3.02 ± 2% 2.93 2.78 2.51 2.52 2.24 ± 4% 1.97 ± 2% 2.16 1.74 1.67 ± 4% 1.58 ± 5% 1.57 ± 2% 1.46 1.65 ± 2% 1.58 ± 2% 1.49 1.36 ± 2% 1.29 ± 5% 1.26 ± 5% 1.21 ± 5% 1.24 ± 2% 1.17 1.10 ± 3% 1.08 ± 3% 1.09 0.95 ± 2% 1.01 ± 2% 1.06 ± 2% 1.10 ± 3% 1.02 ± 2% 0.83 ± 2% 0.87 ± 2% 0.77 ± 5% 0.84 0.89 ± 2% 0.87 ± 2% 0.75 ± 2% 0.79 0.72 ± 7% 0.64 ± 2% 0.73 ± 3% 0.72 0.73 0.74 0.69 ± 10% 0.72 ± 2% 0.62 ± 5% 0.73 0.67 ± 2% 0.53 ± 2% 0.61 ± 2% 0.69 ± 3% 0.61 ± 2% 0.56 ± 2% 0.47 ± 21% 0.51 ± 3% 0.54 ± 2% 0.46 ± 4% 0.55 0.48 ± 3% 0.43 ± 4% 0.54 0.53 0.56 ± 2% 0.72 ± 4% 0.46 ± 3% 0.41 ± 10% 0.41 ± 2% 0.43 ± 10% 0.51 ± 3% 0.48 0.45 ± 2% 0.46 ± 2% 0.38 0.49 ± 2% 0.42 ± 2% 0.46 ± 4% 0.40 ± 2% 0.42 0.36 0.36 ± 3% 0.33 ± 2% 0.32 ± 2% 0.31 ± 3% 0.39 ± 7% 0.44 0.41 ± 7% 0.34 ± 2% 0.30 ± 2% 0.35 ± 3% 0.28 0.27 ± 5% 0.29 ± 2% 0.24 ± 3% 0.32 ± 3% 0.25 ± 4% 0.24 ± 2% 0.28 ± 2% 0.27 ± 4% 0.28 0.35 ± 3% 0.25 0.27 ± 20% 0.29 ± 2% 0.20 ± 8% 0.22 ± 4% 0.32 0.29 ± 2% 0.22 ± 3% 0.20 ± 4% 0.21 ± 2% 0.23 0.22 ± 3% 0.22 ± 3% 0.21 ± 2% 0.21 ± 3% 0.18 ± 3% 0.51 ± 3% 0.22 0.19 ± 4% 0.21 ± 3% 0.20 ± 2% 0.18 ± 2% 0.18 ± 2% 0.23 ± 2% 0.12 ± 4% 0.19 ± 3% 0.23 ± 3% 0.18 ± 2% 0.18 ± 4% 0.15 ± 2% 0.17 ± 4% 0.16 ± 2% 0.13 ± 3% 0.13 0.11 ± 6% 0.08 ± 5% 0.12 ± 9% 0.10 ± 5% 0.11 ± 6% 0.11 0.08 ± 8% 0.11 ± 4% 0.10 ± 4% 0.10 ± 8% 0.13 ± 10% 0.10 ± 4% 0.11 ± 7% 0.08 ± 8% 0.13 0.07 ± 5% 0.07 ± 5% 0.00 0.00 31.18 30.02 49.12 48.68 32.69 32.49 31.56 31.00 20.10 19.82 19.33 17.85 15.28 15.10 14.84 14.85 14.32 13.96 13.18 12.86 12.43 12.27 11.93 11.38 10.30 1.43 ± 3% 0.06 ± 87% 0.15 ± 3% 0.00 4.44 ± 13% 2.90 2.17 1.74 1.39 1.36 ± 2% 1.25 1.08 ± 2% 0.87 ± 2% 1.16 0.83 ± 2% 0.85 1.38 ± 2% 0.89 ± 2% 0.71 ± 2% 0.79 0.82 ± 3% 0.68 ± 2% 0.70 ± 7% 0.68 ± 11% 0.71 0.68 ± 3% 0.66 ± 2% 0.62 0.67 ± 2% 0.65 0.59 0.57 0.65 0.51 ± 2% 0.49 0.52 ± 2% 0.54 ± 3% 0.54 0.45 ± 2% 0.47 ± 3% 0.57 ± 3% 0.42 ± 2% 0.54 ± 2% 0.44 0.39 ± 2% 0.55 0.41 ± 3% 0.40 ± 10% 0.48 0.45 0.40 ± 10% 0.36 ± 4% 0.45 ± 2% 0.42 ± 2% 0.38 ± 5% 0.39 ± 8% 0.41 0.60 ± 4% 0.34 0.36 ± 3% 0.45 ± 4% 0.31 ± 2% 0.33 0.38 ± 6% 0.38 ± 4% 0.31 ± 2% 0.29 ± 3% 0.30 ± 3% 0.27 0.40 ± 2% 0.27 ± 3% 0.31 0.34 ± 3% 0.27 ± 4% 0.29 0.24 ± 3% 0.28 ± 2% 0.24 ± 3% 0.27 0.27 0.27 ± 20% 0.35 ± 4% 0.31 ± 2% 0.24 0.29 ± 3% 0.28 ± 3% 0.27 ± 2% 0.21 ± 5% 0.21 ± 2% 0.18 ± 4% 0.23 0.21 ± 3% 0.26 0.20 ± 4% 0.24 ± 3% 0.20 ± 4% 0.18 ± 3% 0.18 ± 2% 0.20 ± 8% 0.26 ± 5% 0.26 0.20 ± 2% 0.19 ± 3% 0.18 ± 2% 0.22 ± 3% 0.19 ± 3% 0.17 ± 2% 0.20 ± 4% 0.19 ± 3% 0.12 ± 4% 0.23 ± 3% 0.19 ± 4% 0.18 ± 2% 0.18 ± 2% 0.19 ± 6% 0.14 ± 3% 0.16 ± 2% 0.15 ± 4% 0.13 ± 3% 0.15 ± 3% 0.16 ± 2% 0.17 ± 5% 0.14 ± 5% 0.10 ± 4% 0.14 ± 3% 0.14 ± 3% 0.16 ± 5% 0.16 ± 2% 0.15 ± 3% 0.09 ± 4% 0.15 0.13 ± 3% 0.09 0.09 ± 7% 0.12 ± 3% 0.12 0.11 ± 4% 0.10 ± 11% 0.08 0.11 ± 4% 0.11 ± 7% 0.11 ± 7% 0.11 ± 7% 0.11 ± 3% 0.11 ± 4% 0.08 ± 6% 0.10 ± 7% 0.10 0.12 ± 10% 0.11 ± 6% 0.10 ± 4% 0.12 ± 5% 0.09 ± 4% 0.10 0.09 ± 7% 0.09 ± 4% 0.09 ± 4% 0.08 ± 6% 0.09 ± 5% 0.08 ± 5% 0.10 ± 7% 0.07 0.00 0.00 0.00 30.01 0.06 ± 87%

***************** lkp-bdw-ep2: ================= cluster/compiler/ cs-localhost/gc
commit:
6da410d97f b99259a614
6da410d97ffa486e ---------------- fail:runs | :4 9:4 3:4 26:4 1:4 %stddev \ 1930589 87754 1426 3911 385.12 5.792e+08 5.792e+08 5.792e+08 7590488 2095 ± 50% 40809 ± 4% 415.39 211.25 18997 ± 6% 13008831 1.133e+09 8.879e+09 ± 18% 25831364 ± 3% 7.97 17.09 2.34 1037 1.133e+09 33.21 ± 18% 0.96 ± 96% 216.23 523.00 ± 50% 471766 ± 12% 454541 ± 12% 10202 ± 4% 665328 ± 9% 532036 ± 11% 3.597e+12 2.70 9.714e+10 3.173e+11 ± 2% 2.317e+09 1.59 2.871e+13 0.09 4.816e+09 5.29e+12 0.00 ± 27% 50665696 ± 27% 3.418e+12 11.55 ± 5% 4.068e+10 1.81e+13 0.63 7.63e+08 ± 20% 86.73 ± 2% 5.359e+08 ± 9% 83392223 ± 25% 31255 13.66 ± 2% 13.45 ± 2% 11.88 ± 2% 11.67 ± 2% 9.09 ± 2% 7.67 ± 2% 9.02 ± 2% 9.25 ± 2% 8.48 ± 2% 8.14 ± 3% 8.03 ± 3% 6.50 ± 2% 6.11 ± 2% 5.97 ± 2% 5.49 5.35 4.79 4.50 ± 2% 4.24 ± 2% 3.50 ± 3% 2.62 ± 2% 2.88 1.68 2.78 ± 3% 1.84 0.64 ± 3% 1.89 ± 3% 2.04 ± 4% 1.29 ± 2% 1.22 ± 2% 1.32 2.11 ± 2% 1.77 ± 3% 1.49 1.58 ± 4% 1.60 1.50 ± 4% 0.60 ± 2% 1.48 ± 2% 1.41 ± 2% 0.90 1.09 ± 3% 0.81 ± 3% 1.02 ± 3% 0.94 0.72 0.90 ± 3% 0.70 ± 3% 0.71 ± 2% 0.94 ± 3% 0.99 ± 4% 1.07 ± 5% 0.73 ± 3% 1.02 ± 5% 0.68 ± 2% 0.69 ± 4% 0.64 ± 2% 26.69 ± 2% 26.51 ± 2% 25.75 ± 2% 25.20 ± 2% 18.57 ± 2% 18.32 ± 2% 17.95 ± 2% 16.70 ± 2% 14.36 ± 2% 14.21 ± 2% 13.99 ± 2% 14.04 ± 2% 13.60 ± 2% 13.30 ± 2% 0.00 12.53 ± 2% 12.28 ± 2% 11.88 ± 3% 11.73 ± 3% 11.40 ± 2% 0.00 10.89 ± 2% 9.96 ± 2% 0.00 0.00 13.67 ± 2% 13.46 ± 2% 11.88 ± 2% 11.69 ± 2% 11.45 9.11 ± 2% 7.67 ± 2% 9.03 ± 2% 9.25 ± 3% 8.48 ± 2% 8.06 ± 3% 8.15 ± 3% 6.11 ± 2% 5.51 4.80 4.50 ± 2% 4.25 ± 2% 3.66 ± 3% 2.65 ± 2% 2.32 ± 2% 2.38 ± 2% 2.90 2.31 ± 3% 1.68 2.81 ± 3% 2.60 2.42 ± 3% 2.04 ± 4% 1.69 ± 2% 1.29 ± 2% 1.23 ± 2% 1.34 2.12 ± 2% 1.79 ± 3% 1.48 ± 2% 1.24 ± 2% 1.58 ± 4% 1.51 ± 4% 1.48 ± 2% 0.83 ± 3% 1.42 ± 2% 0.73 ± 3% 0.91 1.07 ± 2% 1.10 ± 3% 0.81 ± 3% 0.95 0.90 ± 2% 0.74 0.78 0.49 0.66 ± 2% 0.84 ± 3% 0.71 ± 3% 0.95 ± 3% 0.68 ± 3% 0.57 ± 4% 0.43 ± 7% 0.69 ± 3% 0.43 0.74 ± 3% 0.39 ± 2% 1.07 ± 5% 1.04 ± 5% 0.54 0.57 ± 2% 0.53 ± 2% 0.68 ± 2% 0.57 0.53 0.64 ± 2% 0.56 0.69 ± 5% 0.53 ± 2% 0.39 ± 5% 0.35 ± 4% 0.42 0.61 ± 2% 0.38 ± 6% 0.50 ± 2% 0.40 ± 2% 0.41 ± 3% 0.45 ± 3% 0.45 0.56 ± 4% 0.43 0.30 ± 6% 0.33 ± 3% 0.31 0.20 0.35 0.28 ± 9% 0.21 ± 5% 0.49 0.27 ± 3% 0.09 ± 8% 0.24 ± 5% 0.44 ± 2% 0.26 ± 3% 0.24 0.21 ± 5% 0.15 ± 7% 0.07 ± 12% 0.24 0.18 ± 12% 0.20 ± 4% 0.19 ± 2% 0.17 ± 7% 0.14 ± 10% 0.22 ± 4% 0.29 ± 4% 0.16 ± 4% 0.24 0.22 ± 4% 0.23 ± 3% 0.26 ± 3% 0.21 ± 3% 0.19 ± 6% 0.20 ± 2% 0.18 ± 2% 0.15 ± 7% 0.10 ± 5% 0.19 ± 2% 0.19 ± 4% 0.17 ± 4% 0.17 ± 6% 0.10 ± 7% 0.14 ± 7% 0.13 ± 3% 0.13 ± 3% 0.11 ± 8% 0.10 ± 8% 0.11 ± 4% 0.11 ± 4% 0.09 ± 4% 0.05 0.06 ± 11% 0.07 ± 7% 0.09 ± 4% 0.15 ± 8% 0.11 ± 31% 0.00 0.45 ± 5% 0.00 0.00 26.71 ± 2% 26.51 ± 2% 25.75 ± 2% 25.23 ± 2% 18.58 ± 2% 18.33 ± 2% 17.95 ± 2% 16.73 ± 2% 14.40 ± 2% 14.24 ± 2% 14.05 ± 2% 14.12 ± 2% 13.61 ± 2% 13.31 ± 2% 1.50 ± 2% 0.00 12.28 ± 2% 12.53 ± 2% 11.89 ± 3% 11.73 ± 3% 11.41 ± 2% 10.90 ± 2% 9.97 ± 2% 0.00 0.15 ± 3% 2.25 ± 3% 2.20 1.69 ± 2% 1.22 ± 2% 1.43 ± 2% 1.24 ± 2% 1.01 0.45 1.05 ± 2% 0.74 0.60 ± 2% 0.76 1.35 0.63 ± 2% 0.76 0.67 ± 3% 0.56 ± 3% 0.68 ± 3% 0.42 ± 7% 0.66 0.34 ± 2% 0.59 ± 2% 0.53 ± 3% 0.51 0.64 0.54 0.63 ± 2% 0.40 ± 4% 0.56 0.52 ± 2% 0.20 ± 6% 0.68 ± 5% 0.53 ± 5% 0.38 ± 5% 0.31 ± 2% 0.51 ± 2% 0.26 ± 4% 0.48 ± 2% 0.29 ± 4% 0.44 0.44 0.40 0.25 ± 4% 0.42 0.32 0.28 ± 3% 0.32 ± 2% 0.31 0.49 0.21 ± 4% 0.29 ± 5% 0.09 ± 8% 0.20 ± 7% 0.43 ± 2% 0.35 ± 4% 0.17 ± 2% 0.25 ± 4% 0.08 ± 5% 0.26 ± 4% 0.20 ± 4% 0.12 ± 4% 0.27 0.18 ± 7% 0.14 ± 10% 0.19 ± 6% 0.15 ± 13% 0.20 ± 4% 0.22 ± 5% 0.18 ± 2% 0.20 ± 2% 0.18 ± 3% 0.17 ± 15% 0.29 ± 2% 0.23 ± 3% 0.15 ± 5% 0.16 ± 6% 0.11 ± 3% 0.09 ± 5% 0.09 ± 5% 0.22 ± 5% 0.15 ± 5% 0.15 ± 4% 0.09 ± 4% 0.13 ± 6% 0.15 ± 2% 0.08 ± 5% 0.14 ± 6% 0.13 0.10 0.10 0.09 ± 9% 0.12 ± 3% 0.12 ± 8% 0.06 0.13 ± 3% 0.08 ± 5% 0.09 ± 4% 0.10 ± 9% 0.08 ± 5% 0.13 ± 6% 0.11 ± 7% 0.11 ± 7% 0.18 ± 4% 0.11 ± 3% 0.12 ± 3% 0.09 ± 4% 0.07 0.09 ± 5% 0.08 ± 10% 0.06 ± 11% 0.00 0.00 0.21 ± 7% 0.00 0.38 ± 6% 0.00 0.00 0.00

Disclaimer:
Results for informational design or

Thanks,
Rong Chen
noticed a -6.6% regression of netperf.Throughput_Mbps due to commit:
b99259a61450bb6403bbbbf279f8754a441eae8b ("[PATCH] net/sock: move memory_allocated over to percpu_counter variables")
href="https://github.com/0day-ci/linux/commits/Olof-Johansson/net-sock-move-memory_allocated-over-to-percpu_counter-variables/20180907-095536">https://github.com/0day-ci/linux/commits/Olof-Johansson/net-sock-move-memory_allocated-over-to-percpu_counter-variables/20180907-095536
machine: 8 threads Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz with 8G memory
Netperf is a benchmark that can be use to measure various aspect of networking performance.
href="http://www.netperf.org/netperf/">http://www.netperf.org/netperf/
to that, the commit also has significant impact on the following tests:
--+-----------------------------------------------------------------------+
change | netperf: netperf.Throughput_total_tps -49.3% improvement |
| 88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz with 128G memory |
| cluster=cs-localhost |
| cpufreq_governor=performance |
| ip=ipv4 |
| nr_threads=50% |
| runtime=300s |
| test=UDP_RR |
--+-----------------------------------------------------------------------+
change | netperf: netperf.Throughput_total_tps -13.5% improvement |
| 88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz with 128G memory |
| cluster=cs-localhost |
| cpufreq_governor=performance |
| ip=ipv4 |
| nr_threads=25% |
| runtime=300s |
| test=UDP_RR |
--+-----------------------------------------------------------------------+
--------------------------------------------------------------------------------->
href="https://github.com/intel/lkp-tests.git">https://github.com/intel/lkp-tests.git
bin/lkp install job.yaml # job file is attached in this email
========================================================================
cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/send_size/tbox_group/test/testcase/ucode:
c-7/performance/ipv4/x86_64-rhel-7.2/200%/debian-x86_64-2018-04-03.cgz/300s/5K/lkp-hsw-d01/TCP_SENDFILE/netperf/0x25
(" mlx5e-fixes-2018-09-05")
("net/sock: move memory_allocated over to percpu_counter variables")
b99259a61450bb6403bbbbf279
--------------------------
%reproduction fail:runs
| |
-25% :4 dmesg.RIP:drm_wait_one_vblank[drm]
-25% :4 dmesg.WARNING:at_drivers/gpu/drm/drm_vblank.c:#drm_wait_one_vblank[drm]
25% 1:4 dmesg.WARNING:at_ip_fsnotify/0x
-25% :4 kmsg.drm:drm_atomic_helper_wait_for_dependencies[drm_kms_helper]]*ERROR*[CONNECTOR:#:VGA-#]flip_done_timed_out
-25% :4 kmsg.drm:drm_atomic_helper_wait_for_dependencies[drm_kms_helper]]*ERROR*[CRTC:#:pipe_A]flip_done_timed_out
-25% :4 kmsg.drm:drm_atomic_helper_wait_for_dependencies[drm_kms_helper]]*ERROR*[PLANE:#:primary_A]flip_done_timed_out
-25% :4 kmsg.drm:drm_atomic_helper_wait_for_flip_done[drm_kms_helper]]*ERROR*[CRTC:#:pipe_A]flip_done_timed_out
0% 0:4 perf-profile.children.cycles-pp.schedule_timeout
%change %stddev
| \
-6.6% 7460 netperf.Throughput_Mbps
-6.6% 119362 netperf.Throughput_total_Mbps
-2.1% 1835777 netperf.time.involuntary_context_switches
+1.3% 651.50 netperf.time.percent_of_cpu_this_job_got
+2.1% 1830 netperf.time.system_time
-8.1% 133.19 netperf.time.user_time
-7.7% 1014302 netperf.time.voluntary_context_switches
-6.6% 8.742e+08 netperf.workload
-4.2% 29428 vmstat.system.cs
+77.6% 45981 ± 25% sched_debug.cfs_rq:/.min_vruntime.stddev
+77.6% 45981 ± 25% sched_debug.cfs_rq:/.spread0.stddev
-20.0% 5720 ± 5% slabinfo.anon_vma_chain.active_objs
-21.7% 5804 ± 6% slabinfo.anon_vma_chain.num_objs
-31.7% 26365 ± 22% turbostat.C1E
-0.1 0.08 ± 44% turbostat.C1E%
-24.6% 122121 ± 15% cpuidle.C1.time
-38.8% 1966692 ± 43% cpuidle.C1E.time
-31.8% 26368 ± 22% cpuidle.C1E.usage
-1.4% 1.287e+12 perf-stat.branch-instructions
-0.0 2.97 perf-stat.branch-miss-rate%
-2.5% 3.822e+10 perf-stat.branch-misses
-0.2 0.51 ± 14% perf-stat.cache-miss-rate%
+17.9% 2.477e+10 perf-stat.cache-references
-4.2% 8976073 perf-stat.context-switches
+2.0% 1.33 perf-stat.cpi
-0.0 0.08 perf-stat.dTLB-load-miss-rate%
-15.8% 1.701e+09 perf-stat.dTLB-load-misses
-3.0% 2.135e+12 perf-stat.dTLB-loads
-4.1% 1.325e+12 perf-stat.dTLB-stores
-19.0% 3.908e+08 ± 9% perf-stat.iTLB-loads
-1.9% 6.62e+12 perf-stat.instructions
-1.9% 0.75 perf-stat.ipc
+5.1% 7572 perf-stat.path-length
-1.4 17.04 perf-profile.calltrace.cycles-pp.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-1.4 17.06 perf-profile.calltrace.cycles-pp.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-1.4 16.86 perf-profile.calltrace.cycles-pp.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-1.4 16.80 perf-profile.calltrace.cycles-pp.tcp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64
-1.2 12.24 perf-profile.calltrace.cycles-pp.skb_copy_datagram_iter.tcp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
-1.1 11.57 perf-profile.calltrace.cycles-pp.copy_page_to_iter.skb_copy_datagram_iter.tcp_recvmsg.inet_recvmsg.__sys_recvfrom
-0.9 11.00 ± 2% perf-profile.calltrace.cycles-pp.generic_file_splice_read.splice_direct_to_actor.do_splice_direct.do_sendfile.__x64_sys_sendfile64
-0.9 8.68 perf-profile.calltrace.cycles-pp.copyout.copy_page_to_iter.skb_copy_datagram_iter.tcp_recvmsg.inet_recvmsg
-0.8 8.54 perf-profile.calltrace.cycles-pp.copy_user_enhanced_fast_string.copyout.copy_page_to_iter.skb_copy_datagram_iter.tcp_recvmsg
-0.8 8.73 ± 2% perf-profile.calltrace.cycles-pp.generic_file_read_iter.generic_file_splice_read.splice_direct_to_actor.do_splice_direct.do_sendfile
-0.5 10.21 perf-profile.calltrace.cycles-pp.__tcp_push_pending_frames.do_tcp_sendpages.tcp_sendpage_locked.tcp_sendpage.inet_sendpage
-0.4 10.11 perf-profile.calltrace.cycles-pp.tcp_write_xmit.__tcp_push_pending_frames.do_tcp_sendpages.tcp_sendpage_locked.tcp_sendpage
-0.4 2.59 ± 4% perf-profile.calltrace.cycles-pp.security_file_permission.do_splice_direct.do_sendfile.__x64_sys_sendfile64.do_syscall_64
-0.3 4.60 ± 3% perf-profile.calltrace.cycles-pp.security_file_permission.do_sendfile.__x64_sys_sendfile64.do_syscall_64.entry_SYSCALL_64_after_hwframe
-0.3 0.43 ± 58% perf-profile.calltrace.cycles-pp.__inode_security_revalidate.selinux_file_permission.security_file_permission.do_splice_direct.do_sendfile
-0.3 2.77 ± 4% perf-profile.calltrace.cycles-pp.do_splice_to.splice_direct_to_actor.do_splice_direct.do_sendfile.__x64_sys_sendfile64
-0.2 2.39 ± 2% perf-profile.calltrace.cycles-pp.pagecache_get_page.generic_file_read_iter.generic_file_splice_read.splice_direct_to_actor.do_splice_direct
-0.2 2.09 ± 2% perf-profile.calltrace.cycles-pp.find_get_entry.pagecache_get_page.generic_file_read_iter.generic_file_splice_read.splice_direct_to_actor
-0.2 1.31 ± 8% perf-profile.calltrace.cycles-pp.selinux_file_permission.security_file_permission.do_splice_direct.do_sendfile.__x64_sys_sendfile64
-0.2 0.93 ± 10% perf-profile.calltrace.cycles-pp.__fget_light.do_sendfile.__x64_sys_sendfile64.do_syscall_64.entry_SYSCALL_64_after_hwframe
-0.2 1.68 perf-profile.calltrace.cycles-pp.lock_sock_nested.tcp_sendpage.inet_sendpage.kernel_sendpage.sock_sendpage
-0.2 0.92 ± 3% perf-profile.calltrace.cycles-pp.__kfree_skb.tcp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
-0.2 3.02 ± 2% perf-profile.calltrace.cycles-pp.syscall_return_via_sysret
-0.2 0.56 ± 6% perf-profile.calltrace.cycles-pp.skb_release_data.__kfree_skb.tcp_recvmsg.inet_recvmsg.__sys_recvfrom
-0.2 1.02 ± 5% perf-profile.calltrace.cycles-pp.file_has_perm.security_file_permission.do_splice_direct.do_sendfile.__x64_sys_sendfile64
-0.1 1.05 ± 5% perf-profile.calltrace.cycles-pp.__might_fault.copy_page_to_iter.skb_copy_datagram_iter.tcp_recvmsg.inet_recvmsg
-0.1 1.06 ± 6% perf-profile.calltrace.cycles-pp.file_has_perm.security_file_permission.do_sendfile.__x64_sys_sendfile64.do_syscall_64
-0.1 0.68 perf-profile.calltrace.cycles-pp.___might_sleep.__might_fault.copy_page_to_iter.skb_copy_datagram_iter.tcp_recvmsg
-0.1 0.78 ± 6% perf-profile.calltrace.cycles-pp.atime_needs_update.touch_atime.splice_direct_to_actor.do_splice_direct.do_sendfile
-0.1 0.61 ± 5% perf-profile.calltrace.cycles-pp.sched_clock_cpu.tcp_write_xmit.__tcp_push_pending_frames.do_tcp_sendpages.tcp_sendpage_locked
-0.1 0.96 ± 3% perf-profile.calltrace.cycles-pp.radix_tree_lookup_slot.find_get_entry.pagecache_get_page.generic_file_read_iter.generic_file_splice_read
-0.1 0.64 ± 5% perf-profile.calltrace.cycles-pp._raw_spin_lock_bh.release_sock.tcp_sendpage.inet_sendpage.kernel_sendpage
-0.1 0.55 ± 6% perf-profile.calltrace.cycles-pp.native_sched_clock.sched_clock.sched_clock_cpu.tcp_write_xmit.__tcp_push_pending_frames
-0.1 0.63 ± 4% perf-profile.calltrace.cycles-pp.__tcp_transmit_skb.tcp_write_xmit.__tcp_push_pending_frames.tcp_rcv_established.tcp_v4_do_rcv
-0.1 0.71 ± 5% perf-profile.calltrace.cycles-pp.__radix_tree_lookup.radix_tree_lookup_slot.find_get_entry.pagecache_get_page.generic_file_read_iter
-0.1 0.76 ± 3% perf-profile.calltrace.cycles-pp.__tcp_push_pending_frames.tcp_rcv_established.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish
+0.1 5.71 perf-profile.calltrace.cycles-pp.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip
+0.1 5.49 perf-profile.calltrace.cycles-pp.process_backlog.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq
+0.1 1.04 ± 3% perf-profile.calltrace.cycles-pp.sk_stream_alloc_skb.do_tcp_sendpages.tcp_sendpage_locked.tcp_sendpage.inet_sendpage
+0.4 90.99 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe
+0.5 0.55 ± 6% perf-profile.calltrace.cycles-pp.__release_sock.release_sock.tcp_recvmsg.inet_recvmsg.__sys_recvfrom
+0.6 0.60 ± 6% perf-profile.calltrace.cycles-pp.release_sock.tcp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
+0.6 0.65 ± 18% perf-profile.calltrace.cycles-pp.tcp_data_queue.tcp_rcv_established.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish
+0.7 0.68 ± 2% perf-profile.calltrace.cycles-pp.find_next_bit.cpumask_next.__percpu_counter_sum.__sk_mem_raise_allocated.__sk_mem_schedule
+0.8 0.78 ± 13% perf-profile.calltrace.cycles-pp.percpu_counter_add_batch.__sk_mem_raise_allocated.__sk_mem_schedule.do_tcp_sendpages.tcp_sendpage_locked
+0.8 0.84 perf-profile.calltrace.cycles-pp.cpumask_next.__percpu_counter_sum.__sk_mem_raise_allocated.__sk_mem_schedule.do_tcp_sendpages
+1.7 1.65 ± 4% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__percpu_counter_sum.__sk_mem_raise_allocated.__sk_mem_schedule
+2.0 72.70 perf-profile.calltrace.cycles-pp.__x64_sys_sendfile64.do_syscall_64.entry_SYSCALL_64_after_hwframe
+2.2 70.47 perf-profile.calltrace.cycles-pp.do_sendfile.__x64_sys_sendfile64.do_syscall_64.entry_SYSCALL_64_after_hwframe
+2.8 2.85 ± 5% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__percpu_counter_sum.__sk_mem_raise_allocated.__sk_mem_schedule.do_tcp_sendpages
+3.2 60.70 perf-profile.calltrace.cycles-pp.do_splice_direct.do_sendfile.__x64_sys_sendfile64.do_syscall_64.entry_SYSCALL_64_after_hwframe
+3.6 57.72 perf-profile.calltrace.cycles-pp.splice_direct_to_actor.do_splice_direct.do_sendfile.__x64_sys_sendfile64.do_syscall_64
+4.9 41.77 perf-profile.calltrace.cycles-pp.direct_splice_actor.splice_direct_to_actor.do_splice_direct.do_sendfile.__x64_sys_sendfile64
+5.0 40.88 perf-profile.calltrace.cycles-pp.splice_from_pipe.direct_splice_actor.splice_direct_to_actor.do_splice_direct.do_sendfile
+5.0 40.01 perf-profile.calltrace.cycles-pp.__splice_from_pipe.splice_from_pipe.direct_splice_actor.splice_direct_to_actor.do_splice_direct
+5.0 36.00 perf-profile.calltrace.cycles-pp.pipe_to_sendpage.__splice_from_pipe.splice_from_pipe.direct_splice_actor.splice_direct_to_actor
+5.2 34.87 perf-profile.calltrace.cycles-pp.sock_sendpage.pipe_to_sendpage.__splice_from_pipe.splice_from_pipe.direct_splice_actor
+5.3 33.84 perf-profile.calltrace.cycles-pp.inet_sendpage.kernel_sendpage.sock_sendpage.pipe_to_sendpage.__splice_from_pipe
+5.3 34.14 perf-profile.calltrace.cycles-pp.kernel_sendpage.sock_sendpage.pipe_to_sendpage.__splice_from_pipe.splice_from_pipe
+5.3 32.36 perf-profile.calltrace.cycles-pp.tcp_sendpage.inet_sendpage.kernel_sendpage.sock_sendpage.pipe_to_sendpage
+5.7 27.15 perf-profile.calltrace.cycles-pp.tcp_sendpage_locked.tcp_sendpage.inet_sendpage.kernel_sendpage.sock_sendpage
+5.8 26.38 perf-profile.calltrace.cycles-pp.do_tcp_sendpages.tcp_sendpage_locked.tcp_sendpage.inet_sendpage.kernel_sendpage
+5.9 5.93 ± 4% perf-profile.calltrace.cycles-pp.__percpu_counter_sum.__sk_mem_raise_allocated.__sk_mem_schedule.do_tcp_sendpages.tcp_sendpage_locked
+6.2 7.47 ± 3% perf-profile.calltrace.cycles-pp.__sk_mem_raise_allocated.__sk_mem_schedule.do_tcp_sendpages.tcp_sendpage_locked.tcp_sendpage
+6.3 7.60 ± 3% perf-profile.calltrace.cycles-pp.__sk_mem_schedule.do_tcp_sendpages.tcp_sendpage_locked.tcp_sendpage.inet_sendpage
-1.4 17.04 perf-profile.children.cycles-pp.__sys_recvfrom
-1.4 17.06 perf-profile.children.cycles-pp.__x64_sys_recvfrom
-1.4 16.86 perf-profile.children.cycles-pp.inet_recvmsg
-1.4 16.81 perf-profile.children.cycles-pp.tcp_recvmsg
-1.2 13.82 perf-profile.children.cycles-pp.copy_page_to_iter
-1.2 12.25 perf-profile.children.cycles-pp.skb_copy_datagram_iter
-1.0 11.05 ± 2% perf-profile.children.cycles-pp.generic_file_splice_read
-0.9 8.72 perf-profile.children.cycles-pp.copyout
-0.9 9.23 ± 2% perf-profile.children.cycles-pp.security_file_permission
-0.8 8.60 perf-profile.children.cycles-pp.copy_user_enhanced_fast_string
-0.8 8.90 ± 2% perf-profile.children.cycles-pp.generic_file_read_iter
-0.6 11.03 perf-profile.children.cycles-pp.__tcp_push_pending_frames
-0.5 10.94 perf-profile.children.cycles-pp.tcp_write_xmit
-0.5 5.57 ± 2% perf-profile.children.cycles-pp.selinux_file_permission
-0.3 3.24 ± 2% perf-profile.children.cycles-pp.___might_sleep
-0.3 2.21 ± 5% perf-profile.children.cycles-pp.file_has_perm
-0.3 2.78 ± 4% perf-profile.children.cycles-pp.do_splice_to
-0.3 1.81 ± 2% perf-profile.children.cycles-pp.fsnotify
-0.2 2.45 ± 2% perf-profile.children.cycles-pp.pagecache_get_page
-0.2 3.15 perf-profile.children.cycles-pp.touch_atime
-0.2 1.87 ± 2% perf-profile.children.cycles-pp.lock_sock_nested
-0.2 1.07 ± 2% perf-profile.children.cycles-pp.__kfree_skb
-0.2 2.35 perf-profile.children.cycles-pp.atime_needs_update
-0.2 2.15 ± 3% perf-profile.children.cycles-pp.__might_fault
-0.2 2.13 ± 2% perf-profile.children.cycles-pp.find_get_entry
-0.2 1.00 ± 9% perf-profile.children.cycles-pp.__fget_light
-0.2 3.46 ± 3% perf-profile.children.cycles-pp.syscall_return_via_sysret
-0.2 2.52 ± 2% perf-profile.children.cycles-pp.__inode_security_revalidate
-0.2 1.41 ± 6% perf-profile.children.cycles-pp._raw_spin_lock_bh
-0.2 0.67 ± 5% perf-profile.children.cycles-pp.skb_release_data
-0.2 0.30 ± 28% perf-profile.children.cycles-pp.hrtimer_interrupt
-0.2 0.83 ± 5% perf-profile.children.cycles-pp.__fsnotify_parent
-0.2 0.25 ± 33% perf-profile.children.cycles-pp.__hrtimer_run_queues
-0.2 0.36 ± 24% perf-profile.children.cycles-pp.smp_apic_timer_interrupt
-0.2 0.40 ± 22% perf-profile.children.cycles-pp.apic_timer_interrupt
-0.1 1.34 ± 4% perf-profile.children.cycles-pp.current_time
-0.1 0.18 ± 48% perf-profile.children.cycles-pp.tick_sched_timer
-0.1 1.65 ± 2% perf-profile.children.cycles-pp.__might_sleep
-0.1 0.98 ± 2% perf-profile.children.cycles-pp.radix_tree_lookup_slot
-0.1 0.58 ± 9% perf-profile.children.cycles-pp.nf_hook_slow
-0.1 0.85 ± 4% perf-profile.children.cycles-pp.sched_clock_cpu
-0.1 0.78 ± 3% perf-profile.children.cycles-pp.sched_clock
-0.1 0.75 ± 3% perf-profile.children.cycles-pp.native_sched_clock
-0.1 0.37 ± 8% perf-profile.children.cycles-pp.selinux_ip_postroute
-0.1 0.30 ± 8% perf-profile.children.cycles-pp.selinux_ip_postroute_compat
-0.1 0.38 ± 4% perf-profile.children.cycles-pp.tcp_rate_check_app_limited
-0.1 0.06 ± 6% perf-profile.children.cycles-pp.__copy_skb_header
-0.0 0.15 ± 12% perf-profile.children.cycles-pp.lock_timer_base
-0.0 0.11 ± 18% perf-profile.children.cycles-pp.ip_rcv_core
-0.0 0.17 ± 7% perf-profile.children.cycles-pp.enqueue_to_backlog
-0.0 0.33 ± 6% perf-profile.children.cycles-pp.generic_splice_sendpage
-0.0 0.21 ± 7% perf-profile.children.cycles-pp.iov_iter_pipe
+0.0 0.08 ± 6% perf-profile.children.cycles-pp.__list_del_entry_valid
+0.0 0.08 ± 5% perf-profile.children.cycles-pp.get_nohz_timer_target
+0.0 0.12 ± 6% perf-profile.children.cycles-pp.sk_stream_wait_memory
+0.0 0.12 ± 13% perf-profile.children.cycles-pp.tcp_rearm_rto
+0.1 0.11 ± 14% perf-profile.children.cycles-pp.splice_from_pipe_next
+0.1 1.04 ± 3% perf-profile.children.cycles-pp.sk_stream_alloc_skb
+0.2 0.34 ± 9% perf-profile.children.cycles-pp.tcp_leave_memory_pressure
+0.2 0.58 ± 4% perf-profile.children.cycles-pp.__release_sock
+0.4 3.62 ± 4% perf-profile.children.cycles-pp.tcp_v4_do_rcv
+0.4 0.99 ± 10% perf-profile.children.cycles-pp.tcp_data_queue
+0.4 3.51 ± 5% perf-profile.children.cycles-pp.tcp_rcv_established
+0.4 0.63 ± 15% perf-profile.children.cycles-pp._raw_spin_unlock_irqrestore
+0.5 91.20 perf-profile.children.cycles-pp.do_syscall_64
+0.5 0.60 ± 12% perf-profile.children.cycles-pp.tcp_try_rmem_schedule
+0.8 0.77 ± 2% perf-profile.children.cycles-pp.find_next_bit
+0.9 0.94 ± 10% perf-profile.children.cycles-pp.percpu_counter_add_batch
+1.1 1.14 ± 3% perf-profile.children.cycles-pp.cpumask_next
+1.9 1.92 ± 6% perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
+2.0 72.87 perf-profile.children.cycles-pp.__x64_sys_sendfile64
+2.2 70.62 perf-profile.children.cycles-pp.do_sendfile
+3.2 60.79 perf-profile.children.cycles-pp.do_splice_direct
+3.3 3.54 ± 5% perf-profile.children.cycles-pp._raw_spin_lock_irqsave
+3.6 57.77 perf-profile.children.cycles-pp.splice_direct_to_actor
+5.0 41.83 perf-profile.children.cycles-pp.direct_splice_actor
+5.0 40.97 perf-profile.children.cycles-pp.splice_from_pipe
+5.0 40.13 perf-profile.children.cycles-pp.__splice_from_pipe
+5.0 36.03 perf-profile.children.cycles-pp.pipe_to_sendpage
+5.2 34.97 perf-profile.children.cycles-pp.sock_sendpage
+5.3 33.89 perf-profile.children.cycles-pp.inet_sendpage
+5.3 34.22 perf-profile.children.cycles-pp.kernel_sendpage
+5.3 32.51 perf-profile.children.cycles-pp.tcp_sendpage
+5.7 27.23 perf-profile.children.cycles-pp.tcp_sendpage_locked
+5.8 26.53 perf-profile.children.cycles-pp.do_tcp_sendpages
+6.7 6.67 ± 4% perf-profile.children.cycles-pp.__percpu_counter_sum
+7.0 8.32 ± 3% perf-profile.children.cycles-pp.__sk_mem_raise_allocated
+7.0 8.46 ± 3% perf-profile.children.cycles-pp.__sk_mem_schedule
-0.8 0.40 ± 9% perf-profile.self.cycles-pp.__sk_mem_raise_allocated
-0.8 8.50 perf-profile.self.cycles-pp.copy_user_enhanced_fast_string
-0.3 3.08 perf-profile.self.cycles-pp.___might_sleep
-0.3 2.86 perf-profile.self.cycles-pp.selinux_file_permission
-0.2 1.76 ± 2% perf-profile.self.cycles-pp.fsnotify
-0.2 2.28 perf-profile.self.cycles-pp.generic_file_read_iter
-0.2 2.20 ± 2% perf-profile.self.cycles-pp.do_sendfile
-0.2 3.45 ± 3% perf-profile.self.cycles-pp.syscall_return_via_sysret
-0.2 0.98 ± 9% perf-profile.self.cycles-pp.__fget_light
-0.2 1.45 ± 5% perf-profile.self.cycles-pp.tcp_write_xmit
-0.2 1.02 ± 5% perf-profile.self.cycles-pp.pipe_to_sendpage
-0.2 1.28 ± 6% perf-profile.self.cycles-pp._raw_spin_lock_bh
-0.1 0.74 ± 4% perf-profile.self.cycles-pp.__fsnotify_parent
-0.1 0.98 ± 4% perf-profile.self.cycles-pp.do_syscall_64
-0.1 0.90 ± 5% perf-profile.self.cycles-pp.__x64_sys_sendfile64
-0.1 0.70 ± 6% perf-profile.self.cycles-pp.file_has_perm
-0.1 0.61 ± 7% perf-profile.self.cycles-pp.skb_copy_datagram_iter
-0.1 0.79 ± 6% perf-profile.self.cycles-pp.tcp_current_mss
-0.1 0.04 ± 59% perf-profile.self.cycles-pp.sched_clock
-0.1 0.66 ± 6% perf-profile.self.cycles-pp.skb_release_data
-0.1 0.88 ± 6% perf-profile.self.cycles-pp.page_cache_pipe_buf_confirm
-0.1 0.74 ± 6% perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
-0.1 0.21 ± 8% perf-profile.self.cycles-pp.radix_tree_lookup_slot
-0.1 0.35 ± 4% perf-profile.self.cycles-pp.tcp_rate_check_app_limited
-0.1 1.51 ± 2% perf-profile.self.cycles-pp.__might_sleep
-0.1 0.06 ± 6% perf-profile.self.cycles-pp.__copy_skb_header
-0.1 0.10 ± 15% perf-profile.self.cycles-pp.selinux_ip_postroute_compat
-0.0 0.11 ± 15% perf-profile.self.cycles-pp.ip_rcv_core
-0.0 0.12 ± 8% perf-profile.self.cycles-pp.ip_output
-0.0 0.28 ± 4% perf-profile.self.cycles-pp.generic_splice_sendpage
-0.0 0.10 ± 14% perf-profile.self.cycles-pp.enqueue_to_backlog
-0.0 0.19 ± 2% perf-profile.self.cycles-pp.iov_iter_pipe
-0.0 0.18 ± 7% perf-profile.self.cycles-pp.__ip_queue_xmit
-0.0 0.08 ± 8% perf-profile.self.cycles-pp.sock_def_readable
-0.0 0.08 ± 10% perf-profile.self.cycles-pp.bpf_fd_pass
+0.0 0.15 ± 10% perf-profile.self.cycles-pp.tcp_ack
+0.1 0.10 ± 8% perf-profile.self.cycles-pp.splice_from_pipe_next
+0.2 0.32 ± 9% perf-profile.self.cycles-pp.tcp_leave_memory_pressure
+0.4 0.52 ± 5% perf-profile.self.cycles-pp._raw_spin_unlock_irqrestore
+0.4 0.38 ± 12% perf-profile.self.cycles-pp.cpumask_next
+0.7 0.67 ± 2% perf-profile.self.cycles-pp.find_next_bit
+0.7 0.72 ± 6% perf-profile.self.cycles-pp.percpu_counter_add_batch
+1.4 1.63 ± 6% perf-profile.self.cycles-pp._raw_spin_lock_irqsave
+1.9 1.90 ± 6% perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
+2.1 2.08 ± 3% perf-profile.self.cycles-pp.__percpu_counter_sum

netperf.Throughput_Mbps

-----------------------------------------------------------+
++.+ .++.+++.+ .++. |
+.++.++.++.++.++.++.+++.++.++ ++.++.++.+|
|
|
|
|
|
|
|
O O |
O O O OO OOO OO OO |
|
|
-----------------------------------------------------------+

netperf.Throughput_total_Mbps

+-+----------------------------------------------------------------+
+.+ +.+++.+ +.+ |
: ++.+ +.++.+++.++.+ .+++.++.+ ++.++.++.+|
: : + + |
: : |
: : |
:: |
:: |
:: |
: |
OO O OO O O |
O O OO O OO O |
|
|
+-+----------------------------------------------------------------+

netperf.workload

+-+---------------------------------------------------------------+
+.++ .++.+++ +.+ |
: +.+++.++.++.+++.++.+++.++.++ +.+++.++.+|
: : |
: : |
: : |
: : |
:: |
: |
: |
OO OOO OO O |
OOO O OO |
|
|
+-+---------------------------------------------------------------+

netperf.time.user_time

-----------------------------------------------------------+
+ |
:: |
:: |
+ + : :+. +.+|
.+ .++.+ .++.+ :+ +.++.++.+ :+ +.++. +.+ + + |
+ .++ + +.++ +.+ + +.+ + + |
|
|
|
|
O O OO O O O |
O O OO O O |
|
-----------------------------------------------------------+

netperf.time.system_time

-----------------------------------------------------------+
|
O O OO O OO |
OO O O |
O O O |
|
|
|
|
|
+. +. +.+ .++.++. +. ++. +.++. .+ |
+. + +. + + : + + + + + + ++.+ .++ +.+|
++ + + + +: + |
+ + |
-----------------------------------------------------------+

netperf.time.voluntary_context_switches

+-+--------------------------------------------------------------+
+ + + + + + +. .+ : + .+|
+ + : + + +.++ + + .++.+ :.+ : + .+++ :: +.+++.++ |
+ :.+ + : + + + + + + + |
+ + |
|
|
|
|
OO O OO O |
OO OO O O O |
|
|
|
+-+----O---------------------------------------------------------+

**********************************************************************************
88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz with 128G memory
========================================================================
cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
c-7/performance/ipv4/x86_64-rhel-7.2/50%/debian-x86_64-2018-04-03.cgz/300s/lkp-bdw-ep2/UDP_RR/netperf
(" mlx5e-fixes-2018-09-05")
("net/sock: move memory_allocated over to percpu_counter variables")
b99259a61450bb6403bbbbf279
--------------------------
%reproduction fail:runs
| |
25% 1:4 dmesg.WARNING:at#for_ip_interrupt_entry/0x
-25% :4 dmesg.WARNING:at_ip_native_sched_clock/0x
-149% 5:4 perf-profile.calltrace.cycles-pp.dequeue_entity.dequeue_task_fair.__sched_text_start.schedule.schedule_timeout
-54% 2:4 perf-profile.calltrace.cycles-pp.switch_mm_irqs_off.switch_mm.__sched_text_start.schedule.schedule_timeout
-378% 16:4 perf-profile.children.cycles-pp.schedule_timeout
-12% 0:4 perf-profile.self.cycles-pp.schedule_timeout
%change %stddev
| \
-49.3% 1593635 netperf.Throughput_total_tps
-49.3% 36218 netperf.Throughput_tps
+11.8% 7689 netperf.time.involuntary_context_switches
-10.5% 2501 netperf.time.percent_of_cpu_this_job_got
-6.4% 7194 netperf.time.system_time
-53.7% 336.63 ± 6% netperf.time.user_time
-48.8% 4.781e+08 netperf.time.voluntary_context_switches
-49.3% 4.781e+08 netperf.workload
-10.7% 444.23 pmeter.Average_Active_Power
-43.2% 81.53 pmeter.performance_per_watt
-50.2% 12452 ± 98% proc-vmstat.numa_pte_updates
-1.0% 815119 proc-vmstat.pgfault
+19.6 34.31 ± 2% mpstat.cpu.soft%
-21.0 16.88 ± 5% mpstat.cpu.sys%
-2.4 2.30 ± 6% mpstat.cpu.usr%
+48.8% 345161 ± 22% numa-numastat.node0.numa_hit
-25.1% 343375 ± 24% numa-numastat.node1.local_node
-25.0% 350578 ± 22% numa-numastat.node1.numa_hit
-9.4% 48.00 vmstat.procs.r
-48.6% 6279977 vmstat.system.cs
-5.8% 185888 vmstat.system.in
-49.3% 4.781e+08 softirqs.NET_RX
-23.7% 1595486 ± 3% softirqs.RCU
-13.3% 5915651 softirqs.SCHED
+10.9% 8929857 softirqs.TIMER
+40.3% 1.03e+10 cpuidle.C1.time
-47.4% 9.638e+08 cpuidle.C1.usage
-68.5% 1.243e+08 ± 29% cpuidle.C6.time
-64.4% 199205 ± 18% cpuidle.C6.usage
-47.9% 95056831 cpuidle.POLL.time
-82.4% 6062003 cpuidle.POLL.usage
-14.5% 1708 turbostat.Avg_MHz
-10.8 61.14 turbostat.Busy%
-47.4% 9.638e+08 turbostat.C1
+11.2 38.62 turbostat.C1%
-64.8% 196082 ± 18% turbostat.C6
-1.0 0.45 ± 30% turbostat.C6%
+40.7% 38.28 turbostat.CPU%c1
-13.5% 233.54 turbostat.PkgWatt
+2.1% 15.65 turbostat.RAMWatt
-57.2% 154078 ± 6% sched_debug.cfs_rq:/.MIN_vruntime.avg
-20.3% 2263123 sched_debug.cfs_rq:/.MIN_vruntime.max
-40.1% 552528 ± 3% sched_debug.cfs_rq:/.MIN_vruntime.stddev
-57.2% 154078 ± 6% sched_debug.cfs_rq:/.max_vruntime.avg
-20.3% 2263123 sched_debug.cfs_rq:/.max_vruntime.max
-40.1% 552528 ± 3% sched_debug.cfs_rq:/.max_vruntime.stddev
-19.7% 2251378 sched_debug.cfs_rq:/.min_vruntime.avg
-20.3% 2314293 sched_debug.cfs_rq:/.min_vruntime.max
-18.3% 2232433 sched_debug.cfs_rq:/.min_vruntime.min
-61.3% 11469 ± 21% sched_debug.cfs_rq:/.min_vruntime.stddev
-9.9% 0.52 ± 3% sched_debug.cfs_rq:/.nr_running.avg
-15.4% 8.83 ± 2% sched_debug.cfs_rq:/.runnable_load_avg.avg
-61.3% 11465 ± 21% sched_debug.cfs_rq:/.spread0.stddev
-21.9% 302.54 ± 4% sched_debug.cfs_rq:/.util_est_enqueued.avg
-45.1% 196064 ± 3% sched_debug.cpu.avg_idle.max
-30.2% 41011 ± 2% sched_debug.cpu.avg_idle.stddev
-81.8% 0.08 ± 99% sched_debug.cpu.cpu_load[1].min
-42.6% 1.29 ± 10% sched_debug.cpu.cpu_load[2].min
-21.0% 3.29 ± 4% sched_debug.cpu.cpu_load[3].min
-49.0% 10741980 sched_debug.cpu.nr_switches.avg
-49.7% 10846007 sched_debug.cpu.nr_switches.max
-47.8% 10493962 sched_debug.cpu.nr_switches.min
-79.3% 54430 ± 13% sched_debug.cpu.nr_switches.stddev
-27.2% 4.212e+12 perf-stat.branch-instructions
-0.8 1.96 perf-stat.branch-miss-rate%
-48.2% 8.276e+10 perf-stat.branch-misses
+0.4 1.14 perf-stat.cache-miss-rate%
+11.0% 2.805e+09 perf-stat.cache-misses
-29.1% 2.47e+11 perf-stat.cache-references
-48.8% 1.912e+09 perf-stat.context-switches
+20.0% 2.32 perf-stat.cpi
-17.5% 4.646e+13 perf-stat.cpu-cycles
-43.6% 4.386e+09 ± 15% perf-stat.dTLB-load-misses
-33.9% 5.63e+12 perf-stat.dTLB-loads
+0.0 0.00 ± 7% perf-stat.dTLB-store-miss-rate%
-13.5% 94683885 ± 7% perf-stat.dTLB-store-misses
-48.1% 2.851e+12 perf-stat.dTLB-stores
+3.6 13.65 ± 14% perf-stat.iTLB-load-miss-rate%
-36.1% 5.859e+09 ± 16% perf-stat.iTLB-load-misses
-54.9% 3.696e+10 perf-stat.iTLB-loads
-31.3% 2.002e+13 perf-stat.instructions
-16.7% 0.43 perf-stat.ipc
-1.1% 792718 perf-stat.minor-faults
+108.2% 1.583e+09 perf-stat.node-load-misses
-35.4 57.88 ± 2% perf-stat.node-store-miss-rate%
-40.6% 6.373e+08 ± 2% perf-stat.node-store-misses
+476.1% 4.638e+08 ± 3% perf-stat.node-stores
-1.1% 792720 perf-stat.page-faults
+35.4% 41869 perf-stat.path-length
-7.4 7.94 perf-profile.calltrace.cycles-pp.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-7.3 7.85 perf-profile.calltrace.cycles-pp.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-6.5 6.93 perf-profile.calltrace.cycles-pp.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-6.3 6.82 perf-profile.calltrace.cycles-pp.udp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64
-5.9 2.74 perf-profile.calltrace.cycles-pp.ip_make_skb.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
-5.1 5.43 perf-profile.calltrace.cycles-pp.__skb_recv_udp.udp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
-4.4 4.62 perf-profile.calltrace.cycles-pp.__skb_wait_for_more_packets.__skb_recv_udp.udp_recvmsg.inet_recvmsg.__sys_recvfrom
-4.2 0.88 ± 3% perf-profile.calltrace.cycles-pp.__ip_make_skb.ip_make_skb.udp_sendmsg.sock_sendmsg.__sys_sendto
-3.9 5.70 perf-profile.calltrace.cycles-pp.sock_def_readable.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish
-3.9 5.49 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.sock_def_readable.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv
-3.9 0.65 ± 4% perf-profile.calltrace.cycles-pp.__ip_select_ident.__ip_make_skb.ip_make_skb.udp_sendmsg.sock_sendmsg
-3.9 40.32 perf-profile.calltrace.cycles-pp.secondary_startup_64
-3.9 0.60 ± 5% perf-profile.calltrace.cycles-pp.ip_idents_reserve.__ip_select_ident.__ip_make_skb.ip_make_skb.udp_sendmsg
-3.8 39.95 perf-profile.calltrace.cycles-pp.start_secondary.secondary_startup_64
-3.8 39.94 perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.secondary_startup_64
-3.8 4.06 perf-profile.calltrace.cycles-pp.schedule_timeout.__skb_wait_for_more_packets.__skb_recv_udp.udp_recvmsg.inet_recvmsg
-3.8 39.89 perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
-3.7 5.20 perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_common_lock.sock_def_readable.__udp_enqueue_schedule_skb.udp_queue_rcv_skb
-3.7 3.91 perf-profile.calltrace.cycles-pp.schedule.schedule_timeout.__skb_wait_for_more_packets.__skb_recv_udp.udp_recvmsg
-3.6 3.82 perf-profile.calltrace.cycles-pp.__sched_text_start.schedule.schedule_timeout.__skb_wait_for_more_packets.__skb_recv_udp
-3.6 5.00 perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.sock_def_readable.__udp_enqueue_schedule_skb
-3.6 4.94 perf-profile.calltrace.cycles-pp.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.sock_def_readable
-2.8 3.27 perf-profile.calltrace.cycles-pp.schedule_idle.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
-2.7 3.20 perf-profile.calltrace.cycles-pp.__sched_text_start.schedule_idle.do_idle.cpu_startup_entry.start_secondary
-2.1 2.88 perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock
-1.9 2.69 perf-profile.calltrace.cycles-pp.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common
-1.8 1.84 perf-profile.calltrace.cycles-pp.menu_select.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
-1.7 2.31 perf-profile.calltrace.cycles-pp.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function
-1.6 1.63 perf-profile.calltrace.cycles-pp.dequeue_task_fair.__sched_text_start.schedule.schedule_timeout.__skb_wait_for_more_packets
-1.6 1.69 perf-profile.calltrace.cycles-pp.__ip_append_data.ip_make_skb.udp_sendmsg.sock_sendmsg.__sys_sendto
-1.3 1.19 perf-profile.calltrace.cycles-pp.__entry_SYSCALL_64_trampoline
-1.3 1.20 perf-profile.calltrace.cycles-pp.tick_nohz_get_sleep_length.menu_select.do_idle.cpu_startup_entry.start_secondary
-1.1 1.12 ± 8% perf-profile.calltrace.cycles-pp.__dev_queue_xmit.ip_finish_output2.ip_output.ip_send_skb.udp_send_skb
-1.1 1.18 perf-profile.calltrace.cycles-pp.pick_next_task_fair.__sched_text_start.schedule_idle.do_idle.cpu_startup_entry
-1.0 1.17 perf-profile.calltrace.cycles-pp.sock_alloc_send_pskb.__ip_append_data.ip_make_skb.udp_sendmsg.sock_sendmsg
-1.0 0.96 perf-profile.calltrace.cycles-pp.tick_nohz_next_event.tick_nohz_get_sleep_length.menu_select.do_idle.cpu_startup_entry
-0.8 0.84 ± 10% perf-profile.calltrace.cycles-pp.dev_hard_start_xmit.__dev_queue_xmit.ip_finish_output2.ip_output.ip_send_skb
-0.8 0.78 ± 10% perf-profile.calltrace.cycles-pp.loopback_xmit.dev_hard_start_xmit.__dev_queue_xmit.ip_finish_output2.ip_output
-0.8 0.78 perf-profile.calltrace.cycles-pp.get_next_timer_interrupt.tick_nohz_next_event.tick_nohz_get_sleep_length.menu_select.do_idle
-0.8 0.90 perf-profile.calltrace.cycles-pp.alloc_skb_with_frags.sock_alloc_send_pskb.__ip_append_data.ip_make_skb.udp_sendmsg
-0.7 0.77 perf-profile.calltrace.cycles-pp.syscall_return_via_sysret
-0.7 0.85 perf-profile.calltrace.cycles-pp.__alloc_skb.alloc_skb_with_frags.sock_alloc_send_pskb.__ip_append_data.ip_make_skb
-0.7 0.79 perf-profile.calltrace.cycles-pp.set_next_entity.pick_next_task_fair.__sched_text_start.schedule_idle.do_idle
-0.6 0.65 ± 7% perf-profile.calltrace.cycles-pp.ip_route_output_flow.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
-0.6 0.25 ±100% perf-profile.calltrace.cycles-pp.resched_curr.check_preempt_curr.ttwu_do_wakeup.try_to_wake_up.autoremove_wake_function
-0.6 0.62 ± 7% perf-profile.calltrace.cycles-pp.ip_route_output_key_hash.ip_route_output_flow.udp_sendmsg.sock_sendmsg.__sys_sendto
-0.6 0.86 perf-profile.calltrace.cycles-pp.update_load_avg.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up
-0.6 0.60 ± 7% perf-profile.calltrace.cycles-pp.ip_route_output_key_hash_rcu.ip_route_output_key_hash.ip_route_output_flow.udp_sendmsg.sock_sendmsg
-0.6 0.77 perf-profile.calltrace.cycles-pp.switch_mm_irqs_off.__sched_text_start.schedule_idle.do_idle.cpu_startup_entry
-0.6 0.63 ± 2% perf-profile.calltrace.cycles-pp.switch_mm.__sched_text_start.schedule.schedule_timeout.__skb_wait_for_more_packets
-0.5 0.52 perf-profile.calltrace.cycles-pp.move_addr_to_user.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-0.5 0.59 perf-profile.calltrace.cycles-pp.ttwu_do_wakeup.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock
-0.5 0.65 perf-profile.calltrace.cycles-pp.select_task_rq_fair.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock
-0.4 0.54 ± 2% perf-profile.calltrace.cycles-pp.check_preempt_curr.ttwu_do_wakeup.try_to_wake_up.autoremove_wake_function.__wake_up_common
+2.3 33.16 perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
+2.7 32.39 perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary
+7.3 56.32 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe
+7.5 56.11 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe
+15.2 47.87 perf-profile.calltrace.cycles-pp.__x64_sys_sendto.do_syscall_64.entry_SYSCALL_64_after_hwframe
+15.3 47.77 perf-profile.calltrace.cycles-pp.__sys_sendto.__x64_sys_sendto.do_syscall_64.entry_SYSCALL_64_after_hwframe
+15.7 47.30 perf-profile.calltrace.cycles-pp.sock_sendmsg.__sys_sendto.__x64_sys_sendto.do_syscall_64.entry_SYSCALL_64_after_hwframe
+16.0 47.02 perf-profile.calltrace.cycles-pp.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto.do_syscall_64
+23.1 43.17 perf-profile.calltrace.cycles-pp.udp_send_skb.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
+23.2 43.02 perf-profile.calltrace.cycles-pp.ip_send_skb.udp_send_skb.udp_sendmsg.sock_sendmsg.__sys_sendto
+23.4 42.76 perf-profile.calltrace.cycles-pp.ip_output.ip_send_skb.udp_send_skb.udp_sendmsg.sock_sendmsg
+24.2 42.03 perf-profile.calltrace.cycles-pp.ip_finish_output2.ip_output.ip_send_skb.udp_send_skb.udp_sendmsg
+25.5 40.72 perf-profile.calltrace.cycles-pp.__local_bh_enable_ip.ip_finish_output2.ip_output.ip_send_skb.udp_send_skb
+25.6 40.64 perf-profile.calltrace.cycles-pp.do_softirq.__local_bh_enable_ip.ip_finish_output2.ip_output.ip_send_skb
+25.7 40.52 perf-profile.calltrace.cycles-pp.do_softirq_own_stack.do_softirq.__local_bh_enable_ip.ip_finish_output2.ip_output
+25.7 40.45 perf-profile.calltrace.cycles-pp.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip.ip_finish_output2
+25.9 40.17 perf-profile.calltrace.cycles-pp.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip
+26.0 39.97 perf-profile.calltrace.cycles-pp.process_backlog.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq
+26.4 39.61 perf-profile.calltrace.cycles-pp.__netif_receive_skb_one_core.process_backlog.net_rx_action.__softirqentry_text_start.do_softirq_own_stack
+26.6 39.43 perf-profile.calltrace.cycles-pp.ip_rcv.__netif_receive_skb_one_core.process_backlog.net_rx_action.__softirqentry_text_start
+26.7 39.16 perf-profile.calltrace.cycles-pp.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core.process_backlog.net_rx_action
+26.8 39.07 perf-profile.calltrace.cycles-pp.ip_local_deliver_finish.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core.process_backlog
+27.0 38.90 perf-profile.calltrace.cycles-pp.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core
+27.2 38.58 perf-profile.calltrace.cycles-pp.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv
+27.7 38.01 perf-profile.calltrace.cycles-pp.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver
+30.3 30.29 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__percpu_counter_sum.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb
+30.5 30.48 perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__percpu_counter_sum.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb.udp_queue_rcv_skb
+31.8 31.84 perf-profile.calltrace.cycles-pp.__percpu_counter_sum.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv
+32.0 31.98 perf-profile.calltrace.cycles-pp.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish
-7.4 7.96 perf-profile.children.cycles-pp.__x64_sys_recvfrom
-7.3 7.85 perf-profile.children.cycles-pp.__sys_recvfrom
-6.5 7.12 perf-profile.children.cycles-pp.__sched_text_start
-6.5 6.94 perf-profile.children.cycles-pp.inet_recvmsg
-6.3 6.83 perf-profile.children.cycles-pp.udp_recvmsg
-5.9 2.75 perf-profile.children.cycles-pp.ip_make_skb
-5.1 5.44 perf-profile.children.cycles-pp.__skb_recv_udp
-4.4 4.63 perf-profile.children.cycles-pp.__skb_wait_for_more_packets
-4.2 0.88 ± 3% perf-profile.children.cycles-pp.__ip_make_skb
-3.9 5.71 perf-profile.children.cycles-pp.sock_def_readable
-3.9 5.49 perf-profile.children.cycles-pp.__wake_up_common_lock
-3.9 0.65 ± 4% perf-profile.children.cycles-pp.__ip_select_ident
-3.9 40.32 perf-profile.children.cycles-pp.secondary_startup_64
-3.9 40.32 perf-profile.children.cycles-pp.cpu_startup_entry
-3.9 40.31 perf-profile.children.cycles-pp.do_idle
-3.9 0.60 ± 5% perf-profile.children.cycles-pp.ip_idents_reserve
-3.8 39.95 perf-profile.children.cycles-pp.start_secondary
-3.7 5.20 perf-profile.children.cycles-pp.__wake_up_common
-3.7 3.92 perf-profile.children.cycles-pp.schedule
-3.6 5.01 perf-profile.children.cycles-pp.autoremove_wake_function
-3.6 4.95 perf-profile.children.cycles-pp.try_to_wake_up
-2.9 3.31 perf-profile.children.cycles-pp.schedule_idle
-2.1 2.88 perf-profile.children.cycles-pp.ttwu_do_activate
-1.9 2.69 perf-profile.children.cycles-pp.enqueue_task_fair
-1.8 1.87 perf-profile.children.cycles-pp.menu_select
-1.7 2.36 perf-profile.children.cycles-pp.enqueue_entity
-1.7 1.67 perf-profile.children.cycles-pp.dequeue_task_fair
-1.6 1.69 perf-profile.children.cycles-pp.__ip_append_data
-1.5 1.44 ± 2% perf-profile.children.cycles-pp.__entry_SYSCALL_64_trampoline
-1.5 1.51 perf-profile.children.cycles-pp.dequeue_entity
-1.3 1.61 perf-profile.children.cycles-pp.update_load_avg
-1.3 1.47 perf-profile.children.cycles-pp.pick_next_task_fair
-1.3 1.22 perf-profile.children.cycles-pp.tick_nohz_get_sleep_length
-1.1 1.38 perf-profile.children.cycles-pp.switch_mm_irqs_off
-1.1 1.13 ± 8% perf-profile.children.cycles-pp.__dev_queue_xmit
-1.0 0.97 perf-profile.children.cycles-pp.tick_nohz_next_event
-1.0 1.18 perf-profile.children.cycles-pp.sock_alloc_send_pskb
-0.8 0.89 perf-profile.children.cycles-pp.syscall_return_via_sysret
-0.8 0.84 ± 10% perf-profile.children.cycles-pp.dev_hard_start_xmit
-0.8 0.79 ± 11% perf-profile.children.cycles-pp.loopback_xmit
-0.8 0.79 perf-profile.children.cycles-pp.get_next_timer_interrupt
-0.8 0.70 perf-profile.children.cycles-pp._raw_spin_lock
-0.8 0.90 perf-profile.children.cycles-pp.alloc_skb_with_frags
-0.7 0.85 perf-profile.children.cycles-pp.__alloc_skb
-0.7 0.80 perf-profile.children.cycles-pp.set_next_entity
-0.7 0.67 perf-profile.children.cycles-pp.load_new_mm_cr3
-0.6 0.65 ± 7% perf-profile.children.cycles-pp.ip_route_output_flow
-0.6 0.63 ± 7% perf-profile.children.cycles-pp.ip_route_output_key_hash
-0.6 0.60 ± 7% perf-profile.children.cycles-pp.ip_route_output_key_hash_rcu
-0.6 0.65 perf-profile.children.cycles-pp.switch_mm
-0.6 0.59 perf-profile.children.cycles-pp.update_curr
-0.6 0.54 ± 3% perf-profile.children.cycles-pp.nf_hook_slow
-0.6 0.52 perf-profile.children.cycles-pp.__next_timer_interrupt
-0.5 0.60 ± 2% perf-profile.children.cycles-pp.__update_load_avg_cfs_rq
-0.5 0.45 ± 3% perf-profile.children.cycles-pp.___perf_sw_event
-0.5 0.52 perf-profile.children.cycles-pp.move_addr_to_user
-0.5 0.59 perf-profile.children.cycles-pp.ttwu_do_wakeup
-0.5 0.65 perf-profile.children.cycles-pp.select_task_rq_fair
-0.4 0.57 perf-profile.children.cycles-pp.check_preempt_curr
-0.4 0.41 ± 2% perf-profile.children.cycles-pp.__switch_to_asm
-0.4 0.46 ± 3% perf-profile.children.cycles-pp._copy_to_iter
-0.4 0.37 ± 3% perf-profile.children.cycles-pp.selinux_ip_postroute
-0.4 0.45 ± 3% perf-profile.children.cycles-pp.sk_filter_trim_cap
-0.4 0.50 ± 2% perf-profile.children.cycles-pp.resched_curr
-0.4 0.48 perf-profile.children.cycles-pp.__consume_stateless_skb
-0.4 0.36 ± 2% perf-profile.children.cycles-pp.__switch_to
-0.4 0.42 ± 2% perf-profile.children.cycles-pp.security_sock_rcv_skb
-0.4 0.36 ± 3% perf-profile.children.cycles-pp.update_cfs_group
-0.4 0.29 perf-profile.children.cycles-pp.ip_generic_getfrag
-0.4 0.38 ± 2% perf-profile.children.cycles-pp.tick_nohz_idle_exit
-0.3 0.37 ± 3% perf-profile.children.cycles-pp.finish_task_switch
-0.3 0.39 ± 4% perf-profile.children.cycles-pp.sock_wfree
-0.3 0.40 ± 3% perf-profile.children.cycles-pp.selinux_socket_sock_rcv_skb
-0.3 0.35 ± 12% perf-profile.children.cycles-pp.fib_table_lookup
-0.3 0.39 ± 3% perf-profile.children.cycles-pp.copy_user_generic_unrolled
-0.3 0.30 ± 4% perf-profile.children.cycles-pp.selinux_ip_postroute_compat
-0.3 0.42 perf-profile.children.cycles-pp._raw_spin_lock_bh
-0.3 0.35 ± 3% perf-profile.children.cycles-pp.__slab_free
-0.3 0.23 ± 3% perf-profile.children.cycles-pp._copy_from_iter_full
-0.3 0.33 ± 2% perf-profile.children.cycles-pp.sched_clock_cpu
-0.3 0.40 perf-profile.children.cycles-pp.poll_idle
-0.3 0.34 ± 5% perf-profile.children.cycles-pp.account_entity_enqueue
-0.3 0.30 ± 3% perf-profile.children.cycles-pp.sched_clock
-0.3 0.20 ± 30% perf-profile.children.cycles-pp.netif_rx_internal
-0.3 0.25 ± 4% perf-profile.children.cycles-pp.sockfd_lookup_light
-0.3 0.28 ± 2% perf-profile.children.cycles-pp.native_sched_clock
-0.2 0.21 ± 3% perf-profile.children.cycles-pp.pick_next_task_idle
-0.2 0.30 ± 4% perf-profile.children.cycles-pp.ktime_get
-0.2 0.24 perf-profile.children.cycles-pp.__might_fault
-0.2 0.18 ± 2% perf-profile.children.cycles-pp.hrtimer_next_event_without
-0.2 0.30 ± 2% perf-profile.children.cycles-pp.reweight_entity
-0.2 0.29 perf-profile.children.cycles-pp.prepare_to_wait_exclusive
-0.2 0.33 perf-profile.children.cycles-pp.__update_load_avg_se
-0.2 0.51 perf-profile.children.cycles-pp.find_next_bit
-0.2 0.24 perf-profile.children.cycles-pp.tick_nohz_idle_enter
-0.2 0.19 ± 7% perf-profile.children.cycles-pp.avc_has_perm
-0.2 0.20 ± 4% perf-profile.children.cycles-pp.__fget_light
-0.2 0.21 ± 11% perf-profile.children.cycles-pp.read_tsc
-0.2 0.30 ± 2% perf-profile.children.cycles-pp.update_rq_clock
-0.2 0.27 ± 3% perf-profile.children.cycles-pp.selinux_sock_rcv_skb_compat
-0.2 0.25 perf-profile.children.cycles-pp.native_write_msr
-0.2 0.26 ± 4% perf-profile.children.cycles-pp.sock_def_write_space
-0.2 0.18 ± 2% perf-profile.children.cycles-pp.kmem_cache_alloc_node
-0.2 0.30 ± 2% perf-profile.children.cycles-pp.__kmalloc_reserve
-0.2 0.23 ± 2% perf-profile.children.cycles-pp.update_ts_time_stats
-0.2 0.27 ± 3% perf-profile.children.cycles-pp.copyout
-0.2 0.21 ± 4% perf-profile.children.cycles-pp.ip_local_out
-0.2 0.23 perf-profile.children.cycles-pp.skb_set_owner_w
-0.2 0.18 ± 4% perf-profile.children.cycles-pp.pick_next_entity
-0.2 0.18 ± 4% perf-profile.children.cycles-pp.__ip_local_out
-0.2 0.16 ± 5% perf-profile.children.cycles-pp.validate_xmit_skb
-0.2 0.15 ± 3% perf-profile.children.cycles-pp.enqueue_to_backlog
-0.2 0.15 ± 3% perf-profile.children.cycles-pp.put_prev_task_fair
-0.2 0.23 ± 4% perf-profile.children.cycles-pp.ksize
-0.2 0.28 perf-profile.children.cycles-pp.__kmalloc_node_track_caller
-0.2 0.26 perf-profile.children.cycles-pp.select_idle_sibling
-0.2 0.19 ± 6% perf-profile.children.cycles-pp.ipv4_mtu
-0.1 0.15 ± 3% perf-profile.children.cycles-pp.___might_sleep
-0.1 0.21 ± 2% perf-profile.children.cycles-pp.nr_iowait_cpu
-0.1 0.14 ± 3% perf-profile.children.cycles-pp.__skb_try_recv_from_queue
-0.1 0.12 ± 8% perf-profile.children.cycles-pp.security_socket_sendmsg
-0.1 0.15 ± 5% perf-profile.children.cycles-pp.__udp4_lib_lookup
-0.1 0.10 perf-profile.children.cycles-pp.copyin
-0.1 0.18 ± 3% perf-profile.children.cycles-pp.hrtimer_get_next_event
-0.1 0.11 ± 4% perf-profile.children.cycles-pp.kmem_cache_free
-0.1 0.11 ± 4% perf-profile.children.cycles-pp.entry_SYSCALL_64_stage2
-0.1 0.15 ± 2% perf-profile.children.cycles-pp.move_addr_to_kernel
-0.1 0.14 ± 3% perf-profile.children.cycles-pp.sock_has_perm
-0.1 0.15 ± 5% perf-profile.children.cycles-pp.selinux_parse_skb
-0.1 0.23 ± 3% perf-profile.children.cycles-pp.available_idle_cpu
-0.1 0.12 ± 4% perf-profile.children.cycles-pp._copy_from_user
-0.1 0.15 ± 7% perf-profile.children.cycles-pp.__netif_receive_skb_core
-0.1 0.17 ± 2% perf-profile.children.cycles-pp.__list_del_entry_valid
-0.1 0.08 ± 5% perf-profile.children.cycles-pp.__hrtimer_next_event_base
-0.1 0.11 ± 4% perf-profile.children.cycles-pp.account_entity_dequeue
-0.1 0.20 ± 5% perf-profile.children.cycles-pp.__enqueue_entity
-0.1 0.17 ± 2% perf-profile.children.cycles-pp.skb_release_data
-0.1 0.11 perf-profile.children.cycles-pp._copy_to_user
-0.1 0.09 ± 5% perf-profile.children.cycles-pp.menu_reflect
-0.1 0.10 ± 4% perf-profile.children.cycles-pp.put_prev_entity
-0.1 0.12 ± 3% perf-profile.children.cycles-pp.cpuidle_governor_latency_req
-0.1 0.11 ± 3% perf-profile.children.cycles-pp.copy_user_enhanced_fast_string
-0.1 0.11 ± 4% perf-profile.children.cycles-pp.rcu_idle_exit
-0.1 0.10 ± 4% perf-profile.children.cycles-pp.security_sk_classify_flow
-0.1 0.11 ± 4% perf-profile.children.cycles-pp.sock_recvmsg
-0.1 0.08 ± 5% perf-profile.children.cycles-pp.__get_user_4
-0.1 0.41 ± 2% perf-profile.children.cycles-pp._raw_spin_unlock_irqrestore
-0.1 0.12 ± 8% perf-profile.children.cycles-pp.ip_setup_cork
-0.1 0.09 ± 4% perf-profile.children.cycles-pp.inet_sendmsg
-0.1 0.11 ± 3% perf-profile.children.cycles-pp.__might_sleep
-0.1 0.10 perf-profile.children.cycles-pp.__calc_delta
-0.1 0.09 ± 4% perf-profile.children.cycles-pp.security_socket_recvmsg
-0.1 0.08 ± 5% perf-profile.children.cycles-pp.call_cpuidle
-0.1 0.14 ± 8% perf-profile.children.cycles-pp.__list_add_valid
-0.1 0.04 ± 57% perf-profile.children.cycles-pp.sched_ttwu_pending
-0.1 0.10 ± 4% perf-profile.children.cycles-pp.rcu_eqs_enter
-0.1 0.15 ± 4% perf-profile.children.cycles-pp.activate_task
-0.1 0.10 ± 7% perf-profile.children.cycles-pp.netif_skb_features
-0.1 0.10 ± 4% perf-profile.children.cycles-pp.rb_erase_cached
-0.1 0.07 ± 5% perf-profile.children.cycles-pp.rcu_needs_cpu
-0.1 0.11 ± 4% perf-profile.children.cycles-pp.ip_finish_output
-0.1 0.10 perf-profile.children.cycles-pp.ip_rcv_core
-0.1 0.07 ± 5% perf-profile.children.cycles-pp.selinux_ipv4_output
-0.1 0.07 perf-profile.children.cycles-pp.selinux_sk_getsecid
-0.1 0.05 perf-profile.children.cycles-pp.update_min_vruntime
-0.1 0.03 ±100% perf-profile.children.cycles-pp.ip_send_check
-0.1 0.06 ± 11% perf-profile.children.cycles-pp.compute_score
-0.1 0.04 ± 57% perf-profile.children.cycles-pp.deactivate_task
-0.1 0.05 ± 8% perf-profile.children.cycles-pp.ipv4_pktinfo_prepare
-0.1 0.06 ± 9% perf-profile.children.cycles-pp.dst_release
-0.1 0.03 ±100% perf-profile.children.cycles-pp.__slab_alloc
-0.1 0.06 ± 11% perf-profile.children.cycles-pp.pm_qos_request
-0.1 0.05 perf-profile.children.cycles-pp.udp_rcv
-0.1 0.05 perf-profile.children.cycles-pp.__update_idle_core
-0.1 0.08 ± 6% perf-profile.children.cycles-pp.import_single_range
-0.0 0.05 perf-profile.children.cycles-pp.rcu_eqs_exit
-0.0 0.07 perf-profile.children.cycles-pp.rb_insert_color_cached
-0.0 0.04 ± 58% perf-profile.children.cycles-pp.tick_check_broadcast_expired
-0.0 0.10 ± 4% perf-profile.children.cycles-pp.irq_exit
-0.0 0.05 perf-profile.children.cycles-pp.rebalance_domains
-0.0 0.05 ± 9% perf-profile.children.cycles-pp.receiver_wake_function
+0.1 0.10 ± 5% perf-profile.children.cycles-pp.percpu_counter_add_batch
+0.3 0.28 perf-profile.children.cycles-pp.cpumask_next
+2.3 33.50 perf-profile.children.cycles-pp.cpuidle_enter_state
+2.7 32.70 perf-profile.children.cycles-pp.intel_idle
+7.3 56.39 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
+7.5 56.17 perf-profile.children.cycles-pp.do_syscall_64
+15.2 47.88 perf-profile.children.cycles-pp.__x64_sys_sendto
+15.3 47.77 perf-profile.children.cycles-pp.__sys_sendto
+15.7 47.30 perf-profile.children.cycles-pp.sock_sendmsg
+16.0 47.03 perf-profile.children.cycles-pp.udp_sendmsg
+23.1 43.18 perf-profile.children.cycles-pp.udp_send_skb
+23.2 43.03 perf-profile.children.cycles-pp.ip_send_skb
+23.4 42.77 perf-profile.children.cycles-pp.ip_output
+24.2 42.05 perf-profile.children.cycles-pp.ip_finish_output2
+25.5 40.76 perf-profile.children.cycles-pp.__local_bh_enable_ip
+25.6 40.66 perf-profile.children.cycles-pp.do_softirq
+25.7 40.52 perf-profile.children.cycles-pp.do_softirq_own_stack
+25.7 40.54 perf-profile.children.cycles-pp.__softirqentry_text_start
+25.9 40.18 perf-profile.children.cycles-pp.net_rx_action
+26.0 39.98 perf-profile.children.cycles-pp.process_backlog
+26.4 39.61 perf-profile.children.cycles-pp.__netif_receive_skb_one_core
+26.6 39.44 perf-profile.children.cycles-pp.ip_rcv
+26.7 39.16 perf-profile.children.cycles-pp.ip_local_deliver
+26.8 39.08 perf-profile.children.cycles-pp.ip_local_deliver_finish
+27.0 38.90 perf-profile.children.cycles-pp.__udp4_lib_rcv
+27.2 38.59 perf-profile.children.cycles-pp.udp_queue_rcv_skb
+27.7 38.02 perf-profile.children.cycles-pp.__udp_enqueue_schedule_skb
+29.9 31.29 perf-profile.children.cycles-pp._raw_spin_lock_irqsave
+30.2 30.31 perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
+31.8 31.98 perf-profile.children.cycles-pp.__sk_mem_raise_allocated
+31.9 31.87 perf-profile.children.cycles-pp.__percpu_counter_sum
-3.8 0.60 ± 5% perf-profile.self.cycles-pp.ip_idents_reserve
-1.5 1.39 perf-profile.self.cycles-pp.__entry_SYSCALL_64_trampoline
-1.0 1.20 perf-profile.self.cycles-pp.__sched_text_start
-0.8 0.89 perf-profile.self.cycles-pp.syscall_return_via_sysret
-0.7 0.67 perf-profile.self.cycles-pp._raw_spin_lock
-0.7 0.67 perf-profile.self.cycles-pp.load_new_mm_cr3
-0.6 0.67 perf-profile.self.cycles-pp.update_load_avg
-0.5 0.59 ± 2% perf-profile.self.cycles-pp.__update_load_avg_cfs_rq
-0.5 0.41 ± 3% perf-profile.self.cycles-pp.___perf_sw_event
-0.4 0.71 perf-profile.self.cycles-pp.switch_mm_irqs_off
-0.4 0.41 ± 2% perf-profile.self.cycles-pp.__switch_to_asm
-0.4 0.43 ± 2% perf-profile.self.cycles-pp.update_curr
-0.4 0.98 perf-profile.self.cycles-pp._raw_spin_lock_irqsave
-0.4 0.49 perf-profile.self.cycles-pp.menu_select
-0.4 0.35 ± 2% perf-profile.self.cycles-pp.__switch_to
-0.4 0.43 perf-profile.self.cycles-pp.do_idle
-0.4 0.46 ± 2% perf-profile.self.cycles-pp.resched_curr
-0.4 0.32 ± 3% perf-profile.self.cycles-pp.udp_sendmsg
-0.3 0.35 ± 3% perf-profile.self.cycles-pp.update_cfs_group
-0.3 0.35 ± 12% perf-profile.self.cycles-pp.fib_table_lookup
-0.3 0.41 perf-profile.self.cycles-pp._raw_spin_lock_bh
-0.3 0.38 ± 2% perf-profile.self.cycles-pp.copy_user_generic_unrolled
-0.3 0.35 ± 3% perf-profile.self.cycles-pp.__slab_free
-0.3 0.32 ± 2% perf-profile.self.cycles-pp.pick_next_task_fair
-0.3 0.38 perf-profile.self.cycles-pp.select_task_rq_fair
-0.3 0.36 perf-profile.self.cycles-pp.set_next_entity
-0.3 0.32 perf-profile.self.cycles-pp.udp_recvmsg
-0.3 0.30 ± 2% perf-profile.self.cycles-pp.finish_task_switch
-0.3 0.39 perf-profile.self.cycles-pp.enqueue_entity
-0.3 0.26 ± 3% perf-profile.self.cycles-pp.do_syscall_64
-0.3 0.24 ± 3% perf-profile.self.cycles-pp.ip_route_output_key_hash_rcu
-0.2 0.27 ± 2% perf-profile.self.cycles-pp.native_sched_clock
-0.2 0.29 ± 2% perf-profile.self.cycles-pp.dequeue_entity
-0.2 0.30 ± 2% perf-profile.self.cycles-pp.reweight_entity
-0.2 0.21 ± 7% perf-profile.self.cycles-pp.__ip_append_data
-0.2 0.24 perf-profile.self.cycles-pp.__next_timer_interrupt
-0.2 0.34 ± 3% perf-profile.self.cycles-pp.poll_idle
-0.2 0.19 ± 3% perf-profile.self.cycles-pp.__skb_wait_for_more_packets
-0.2 0.32 perf-profile.self.cycles-pp.__update_load_avg_se
-0.2 0.21 ± 2% perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
-0.2 0.16 ± 6% perf-profile.self.cycles-pp.ip_output
-0.2 0.33 ± 2% perf-profile.self.cycles-pp.enqueue_task_fair
-0.2 0.20 ± 4% perf-profile.self.cycles-pp.__fget_light
-0.2 0.18 ± 6% perf-profile.self.cycles-pp.avc_has_perm
-0.2 0.28 perf-profile.self.cycles-pp.move_addr_to_user
-0.2 0.24 ± 2% perf-profile.self.cycles-pp.native_write_msr
-0.2 0.20 ± 13% perf-profile.self.cycles-pp.read_tsc
-0.2 0.16 ± 4% perf-profile.self.cycles-pp.__skb_recv_udp
-0.2 0.25 ± 5% perf-profile.self.cycles-pp.sock_def_write_space
-0.2 0.23 ± 2% perf-profile.self.cycles-pp.account_entity_enqueue
-0.2 0.19 ± 2% perf-profile.self.cycles-pp.dequeue_task_fair
-0.2 0.20 ± 3% perf-profile.self.cycles-pp.ip_finish_output2
-0.2 0.23 perf-profile.self.cycles-pp.skb_set_owner_w
-0.2 0.42 perf-profile.self.cycles-pp.find_next_bit
-0.2 0.17 ± 3% perf-profile.self.cycles-pp.__ip_make_skb
-0.2 0.19 ± 2% perf-profile.self.cycles-pp.process_backlog
-0.2 0.28 perf-profile.self.cycles-pp._raw_spin_unlock_irqrestore
-0.2 0.15 ± 2% perf-profile.self.cycles-pp.__alloc_skb
-0.2 0.17 ± 4% perf-profile.self.cycles-pp.pick_next_entity
-0.2 0.22 ± 3% perf-profile.self.cycles-pp.ksize
-0.2 0.22 ± 3% perf-profile.self.cycles-pp.__softirqentry_text_start
-0.2 0.15 ± 2% perf-profile.self.cycles-pp.__sys_sendto
-0.2 0.14 ± 8% perf-profile.self.cycles-pp._copy_to_iter
-0.1 0.15 perf-profile.self.cycles-pp.___might_sleep
-0.1 0.12 ± 5% perf-profile.self.cycles-pp.sock_wfree
-0.1 0.26 ± 3% perf-profile.self.cycles-pp.__udp_enqueue_schedule_skb
-0.1 0.13 ± 3% perf-profile.self.cycles-pp.__sys_recvfrom
-0.1 0.17 ± 6% perf-profile.self.cycles-pp.ipv4_mtu
-0.1 0.21 ± 2% perf-profile.self.cycles-pp.nr_iowait_cpu
-0.1 0.13 ± 3% perf-profile.self.cycles-pp.sock_has_perm
-0.1 0.16 ± 5% perf-profile.self.cycles-pp.net_rx_action
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.entry_SYSCALL_64_stage2
-0.1 0.15 ± 4% perf-profile.self.cycles-pp.loopback_xmit
-0.1 0.11 ± 4% perf-profile.self.cycles-pp.inet_recvmsg
-0.1 0.14 ± 3% perf-profile.self.cycles-pp.do_softirq
-0.1 0.15 ± 5% perf-profile.self.cycles-pp.selinux_parse_skb
-0.1 0.15 ± 5% perf-profile.self.cycles-pp.__netif_receive_skb_core
-0.1 0.23 perf-profile.self.cycles-pp.available_idle_cpu
-0.1 0.19 ± 3% perf-profile.self.cycles-pp.__enqueue_entity
-0.1 0.12 ± 3% perf-profile.self.cycles-pp.selinux_socket_sock_rcv_skb
-0.1 0.17 ± 4% perf-profile.self.cycles-pp.__list_del_entry_valid
-0.1 0.17 ± 3% perf-profile.self.cycles-pp.skb_release_data
-0.1 0.16 ± 4% perf-profile.self.cycles-pp.__udp4_lib_rcv
-0.1 0.10 ± 5% perf-profile.self.cycles-pp.account_entity_dequeue
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.kmem_cache_alloc_node
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.tick_nohz_next_event
-0.1 0.12 ± 4% perf-profile.self.cycles-pp.udp_send_skb
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.copy_user_enhanced_fast_string
-0.1 0.16 ± 4% perf-profile.self.cycles-pp.try_to_wake_up
-0.1 0.09 ± 4% perf-profile.self.cycles-pp.put_prev_task_fair
-0.1 0.14 ± 3% perf-profile.self.cycles-pp.__wake_up_common
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.__x64_sys_recvfrom
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.selinux_ip_postroute_compat
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.__get_user_4
-0.1 0.11 ± 4% perf-profile.self.cycles-pp.__dev_queue_xmit
-0.1 0.16 ± 2% perf-profile.self.cycles-pp.update_rq_clock
-0.1 0.16 ± 4% perf-profile.self.cycles-pp.__kmalloc_node_track_caller
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.__local_bh_enable_ip
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.enqueue_to_backlog
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.call_cpuidle
-0.1 0.13 ± 9% perf-profile.self.cycles-pp.__list_add_valid
-0.1 0.10 perf-profile.self.cycles-pp.__calc_delta
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.inet_sendmsg
-0.1 0.11 ± 6% perf-profile.self.cycles-pp.__x64_sys_sendto
-0.1 0.10 ± 4% perf-profile.self.cycles-pp.rcu_eqs_enter
-0.1 0.04 ± 57% perf-profile.self.cycles-pp.sched_ttwu_pending
-0.1 0.15 ± 4% perf-profile.self.cycles-pp.activate_task
-0.1 0.11 ± 4% perf-profile.self.cycles-pp.check_preempt_curr
-0.1 0.10 perf-profile.self.cycles-pp.__might_sleep
-0.1 0.09 ± 4% perf-profile.self.cycles-pp.__udp4_lib_lookup
-0.1 0.11 ± 18% perf-profile.self.cycles-pp.ktime_get
-0.1 0.06 ± 15% perf-profile.self.cycles-pp.validate_xmit_skb
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.selinux_sock_rcv_skb_compat
-0.1 0.07 ± 5% perf-profile.self.cycles-pp.selinux_ip_postroute
-0.1 0.05 perf-profile.self.cycles-pp._copy_from_iter_full
-0.1 0.07 ± 10% perf-profile.self.cycles-pp.rcu_needs_cpu
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.prepare_to_wait_exclusive
-0.1 0.09 ± 4% perf-profile.self.cycles-pp.ip_local_deliver_finish
-0.1 0.07 ± 10% perf-profile.self.cycles-pp.netif_skb_features
-0.1 0.03 ±100% perf-profile.self.cycles-pp.put_prev_entity
-0.1 0.07 ± 7% perf-profile.self.cycles-pp.__might_fault
-0.1 0.07 ± 6% perf-profile.self.cycles-pp.menu_reflect
-0.1 0.09 perf-profile.self.cycles-pp.rb_erase_cached
-0.1 0.10 ± 5% perf-profile.self.cycles-pp.ip_rcv_core
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.ip_rcv
-0.1 0.03 ±100% perf-profile.self.cycles-pp.udp_rcv
-0.1 0.08 ± 5% perf-profile.self.cycles-pp.ip_local_deliver
-0.1 0.07 ± 6% perf-profile.self.cycles-pp.__hrtimer_next_event_base
-0.1 0.03 ±100% perf-profile.self.cycles-pp.selinux_ipv4_output
-0.1 0.03 ±100% perf-profile.self.cycles-pp.deactivate_task
-0.1 0.06 ± 6% perf-profile.self.cycles-pp.rcu_idle_exit
-0.1 0.06 perf-profile.self.cycles-pp.selinux_sk_getsecid
-0.1 0.05 perf-profile.self.cycles-pp.ip_make_skb
-0.1 0.04 ± 57% perf-profile.self.cycles-pp.__update_idle_core
-0.1 0.03 ±100% perf-profile.self.cycles-pp.ip_send_check
-0.1 0.06 ± 11% perf-profile.self.cycles-pp.pm_qos_request
-0.1 0.06 ± 11% perf-profile.self.cycles-pp.compute_score
-0.1 0.05 ± 9% perf-profile.self.cycles-pp.sock_sendmsg
-0.1 0.06 ± 9% perf-profile.self.cycles-pp.ip_generic_getfrag
-0.1 0.06 perf-profile.self.cycles-pp.switch_mm
-0.1 0.06 ± 9% perf-profile.self.cycles-pp.nf_hook_slow
-0.1 0.03 ±100% perf-profile.self.cycles-pp.alloc_skb_with_frags
-0.1 0.05 perf-profile.self.cycles-pp.__skb_try_recv_from_queue
-0.1 0.05 perf-profile.self.cycles-pp.dst_release
-0.0 0.07 perf-profile.self.cycles-pp.import_single_range
-0.0 0.06 perf-profile.self.cycles-pp.rb_insert_color_cached
-0.0 0.05 ± 8% perf-profile.self.cycles-pp.sock_alloc_send_pskb
-0.0 0.08 ± 6% perf-profile.self.cycles-pp.cpuidle_enter_state
-0.0 0.05 perf-profile.self.cycles-pp.ip_finish_output
-0.0 0.06 perf-profile.self.cycles-pp.ip_send_skb
-0.0 0.05 perf-profile.self.cycles-pp.schedule_idle
-0.0 0.05 perf-profile.self.cycles-pp.get_next_timer_interrupt
-0.0 0.06 ± 9% perf-profile.self.cycles-pp.udp_queue_rcv_skb
-0.0 0.04 ± 58% perf-profile.self.cycles-pp.tick_check_broadcast_expired
-0.0 0.05 perf-profile.self.cycles-pp.dev_hard_start_xmit
-0.0 0.05 perf-profile.self.cycles-pp.ip_setup_cork
-0.0 0.07 ± 6% perf-profile.self.cycles-pp.do_softirq_own_stack
-0.0 0.05 perf-profile.self.cycles-pp.receiver_wake_function
+0.1 0.10 ± 5% perf-profile.self.cycles-pp.percpu_counter_add_batch
+0.1 0.10 ± 4% perf-profile.self.cycles-pp.cpumask_next
+1.0 1.01 perf-profile.self.cycles-pp.__percpu_counter_sum
+2.7 32.69 perf-profile.self.cycles-pp.intel_idle
+30.2 30.31 perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
**********************************************************************************
88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz with 128G memory
========================================================================
cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
c-7/performance/ipv4/x86_64-rhel-7.2/25%/debian-x86_64-2018-04-03.cgz/300s/lkp-bdw-ep2/UDP_RR/netperf
(" mlx5e-fixes-2018-09-05")
("net/sock: move memory_allocated over to percpu_counter variables")
b99259a61450bb6403bbbbf279
--------------------------
%reproduction fail:runs
| |
25% 1:4 dmesg.WARNING:at#for_ip_interrupt_entry/0x
-43% 7:4 perf-profile.calltrace.cycles-pp.dequeue_entity.dequeue_task_fair.__sched_text_start.schedule.schedule_timeout
-15% 3:4 perf-profile.calltrace.cycles-pp.switch_mm_irqs_off.switch_mm.__sched_text_start.schedule.schedule_timeout
-109% 21:4 perf-profile.children.cycles-pp.schedule_timeout
-11% 1:4 perf-profile.self.cycles-pp.schedule_timeout
%change %stddev
| \
-13.5% 1670226 netperf.Throughput_total_tps
-13.5% 75919 netperf.Throughput_tps
-3.9% 1371 ± 2% netperf.time.percent_of_cpu_this_job_got
-2.6% 3809 ± 2% netperf.time.system_time
-17.0% 319.68 ± 4% netperf.time.user_time
-13.5% 5.011e+08 netperf.time.voluntary_context_switches
-13.5% 5.011e+08 netperf.workload
-12.5% 5.068e+08 softirqs.NET_RX
-13.1% 6595363 vmstat.system.cs
+740.9% 17617 ± 87% numa-meminfo.node0.Shmem
-14.3% 34971 ± 11% numa-meminfo.node1.SReclaimable
-1.8% 408.09 pmeter.Average_Active_Power
-11.9% 186.02 pmeter.performance_per_watt
+8.2% 20555 ± 3% sched_debug.cpu.load.avg
-13.3% 11273636 sched_debug.cpu.nr_switches.avg
-12.8% 9.877e+08 cpuidle.C1.usage
-43.5% 5.013e+09 ± 54% cpuidle.C6.time
-27.4% 18752863 cpuidle.POLL.usage
+2.5 10.43 mpstat.cpu.soft%
-2.4 14.66 mpstat.cpu.sys%
-0.3 2.00 ± 2% mpstat.cpu.usr%
-3.8% 998.50 turbostat.Avg_MHz
-12.8% 9.877e+08 turbostat.C1
-14.4 18.80 ± 55% turbostat.C6%
+201.3% 2.88 ± 30% turbostat.CPU%c3
-2.2% 211.46 turbostat.PkgWatt
+742.2% 4404 ± 87% numa-vmstat.node0.nr_shmem
+27.2% 600222 ± 6% numa-vmstat.node0.numa_hit
+28.6% 584644 ± 7% numa-vmstat.node0.numa_local
-14.3% 8743 ± 11% numa-vmstat.node1.nr_slab_reclaimable
-20.2% 530868 ± 7% numa-vmstat.node1.numa_hit
-25.5% 396167 ± 9% numa-vmstat.node1.numa_local
-6.9% 3.348e+12 perf-stat.branch-instructions
-0.2 2.54 perf-stat.branch-miss-rate%
-12.4% 8.509e+10 perf-stat.branch-misses
-14.9% 2.701e+11 perf-stat.cache-references
-13.3% 2.01e+09 perf-stat.context-switches
+3.4% 1.64 perf-stat.cpi
-4.5% 2.742e+13 perf-stat.cpu-cycles
-0.0 0.09 perf-stat.dTLB-load-miss-rate%
-10.7% 4.302e+09 perf-stat.dTLB-load-misses
-8.4% 4.846e+12 perf-stat.dTLB-loads
+0.0 0.00 ± 9% perf-stat.dTLB-store-miss-rate%
+87.6% 95071493 ± 9% perf-stat.dTLB-store-misses
-11.6% 3.023e+12 perf-stat.dTLB-stores
+1.9 13.47 ± 7% perf-stat.iTLB-load-miss-rate%
-16.8% 3.384e+10 ± 2% perf-stat.iTLB-loads
-7.6% 1.673e+13 perf-stat.instructions
-3.3% 0.61 perf-stat.ipc
+87.8% 1.433e+09 ± 3% perf-stat.node-load-misses
-27.7 59.01 perf-stat.node-store-miss-rate%
+17.5% 6.296e+08 perf-stat.node-store-misses
+424.4% 4.373e+08 perf-stat.node-stores
+6.8% 33385 perf-stat.path-length
-2.4 11.21 ± 6% perf-profile.calltrace.cycles-pp.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-2.4 11.05 ± 6% perf-profile.calltrace.cycles-pp.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-2.1 9.73 ± 6% perf-profile.calltrace.cycles-pp.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-2.1 9.58 ± 6% perf-profile.calltrace.cycles-pp.udp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64
-1.6 7.45 ± 7% perf-profile.calltrace.cycles-pp.__skb_recv_udp.udp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
-1.4 6.27 ± 7% perf-profile.calltrace.cycles-pp.__skb_wait_for_more_packets.__skb_recv_udp.udp_recvmsg.inet_recvmsg.__sys_recvfrom
-1.3 7.69 ± 6% perf-profile.calltrace.cycles-pp.__wake_up_common_lock.sock_def_readable.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv
-1.3 7.99 ± 6% perf-profile.calltrace.cycles-pp.sock_def_readable.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish
-1.2 7.24 ± 7% perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_common_lock.sock_def_readable.__udp_enqueue_schedule_skb.udp_queue_rcv_skb
-1.1 6.99 ± 7% perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.sock_def_readable.__udp_enqueue_schedule_skb
-1.1 6.89 ± 7% perf-profile.calltrace.cycles-pp.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.sock_def_readable
-1.1 5.41 ± 7% perf-profile.calltrace.cycles-pp.schedule_timeout.__skb_wait_for_more_packets.__skb_recv_udp.udp_recvmsg.inet_recvmsg
-1.0 5.12 ± 7% perf-profile.calltrace.cycles-pp.schedule.schedule_timeout.__skb_wait_for_more_packets.__skb_recv_udp.udp_recvmsg
-1.0 5.02 ± 7% perf-profile.calltrace.cycles-pp.__sched_text_start.schedule.schedule_timeout.__skb_wait_for_more_packets.__skb_recv_udp
-0.9 4.60 ± 5% perf-profile.calltrace.cycles-pp.schedule_idle.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
-0.9 4.48 ± 5% perf-profile.calltrace.cycles-pp.__sched_text_start.schedule_idle.do_idle.cpu_startup_entry.start_secondary
-0.8 3.96 ± 6% perf-profile.calltrace.cycles-pp.ip_make_skb.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
-0.7 3.85 ± 7% perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock
-0.6 3.62 ± 8% perf-profile.calltrace.cycles-pp.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common
-0.5 2.99 ± 8% perf-profile.calltrace.cycles-pp.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function
-0.5 2.16 ± 7% perf-profile.calltrace.cycles-pp.dequeue_task_fair.__sched_text_start.schedule.schedule_timeout.__skb_wait_for_more_packets
-0.4 2.46 ± 6% perf-profile.calltrace.cycles-pp.__ip_append_data.ip_make_skb.udp_sendmsg.sock_sendmsg.__sys_sendto
-0.4 1.29 ± 4% perf-profile.calltrace.cycles-pp.__ip_make_skb.ip_make_skb.udp_sendmsg.sock_sendmsg.__sys_sendto
-0.4 2.42 ± 7% perf-profile.calltrace.cycles-pp.menu_select.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
-0.4 1.48 ± 6% perf-profile.calltrace.cycles-pp.pick_next_task_fair.__sched_text_start.schedule_idle.do_idle.cpu_startup_entry
-0.4 0.28 ±100% perf-profile.calltrace.cycles-pp.__next_timer_interrupt.get_next_timer_interrupt.tick_nohz_next_event.tick_nohz_get_sleep_length.menu_select
-0.3 1.56 ± 7% perf-profile.calltrace.cycles-pp.__entry_SYSCALL_64_trampoline
-0.3 1.73 ± 8% perf-profile.calltrace.cycles-pp.__dev_queue_xmit.ip_finish_output2.ip_output.ip_send_skb.udp_send_skb
-0.3 1.00 ± 4% perf-profile.calltrace.cycles-pp.__ip_select_ident.__ip_make_skb.ip_make_skb.udp_sendmsg.sock_sendmsg
-0.3 0.94 ± 4% perf-profile.calltrace.cycles-pp.ip_idents_reserve.__ip_select_ident.__ip_make_skb.ip_make_skb.udp_sendmsg
-0.3 1.04 ± 6% perf-profile.calltrace.cycles-pp.set_next_entity.pick_next_task_fair.__sched_text_start.schedule_idle.do_idle
-0.3 1.83 ± 6% perf-profile.calltrace.cycles-pp.sock_alloc_send_pskb.__ip_append_data.ip_make_skb.udp_sendmsg.sock_sendmsg
-0.3 1.50 ± 8% perf-profile.calltrace.cycles-pp.tick_nohz_get_sleep_length.menu_select.do_idle.cpu_startup_entry.start_secondary
-0.3 1.22 ± 5% perf-profile.calltrace.cycles-pp.syscall_return_via_sysret
-0.2 1.36 ± 7% perf-profile.calltrace.cycles-pp.dev_hard_start_xmit.__dev_queue_xmit.ip_finish_output2.ip_output.ip_send_skb
-0.2 1.40 ± 4% perf-profile.calltrace.cycles-pp.switch_mm_irqs_off.__sched_text_start.schedule_idle.do_idle.cpu_startup_entry
-0.2 1.30 ± 7% perf-profile.calltrace.cycles-pp.loopback_xmit.dev_hard_start_xmit.__dev_queue_xmit.ip_finish_output2.ip_output
-0.2 0.41 ± 58% perf-profile.calltrace.cycles-pp.selinux_socket_sock_rcv_skb.security_sock_rcv_skb.sk_filter_trim_cap.udp_queue_rcv_skb.__udp4_lib_rcv
-0.2 1.29 ± 7% perf-profile.calltrace.cycles-pp.alloc_skb_with_frags.sock_alloc_send_pskb.__ip_append_data.ip_make_skb.udp_sendmsg
-0.2 1.23 ± 7% perf-profile.calltrace.cycles-pp.__alloc_skb.alloc_skb_with_frags.sock_alloc_send_pskb.__ip_append_data.ip_make_skb
-0.2 0.73 ± 4% perf-profile.calltrace.cycles-pp._copy_to_iter.udp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
-0.2 0.93 ± 8% perf-profile.calltrace.cycles-pp.get_next_timer_interrupt.tick_nohz_next_event.tick_nohz_get_sleep_length.menu_select.do_idle
-0.2 0.66 ± 4% perf-profile.calltrace.cycles-pp.sock_wfree.loopback_xmit.dev_hard_start_xmit.__dev_queue_xmit.ip_finish_output2
-0.2 0.87 ± 7% perf-profile.calltrace.cycles-pp.switch_mm.__sched_text_start.schedule.schedule_timeout.__skb_wait_for_more_packets
-0.2 0.79 ± 6% perf-profile.calltrace.cycles-pp.move_addr_to_user.__sys_recvfrom.__x64_sys_recvfrom.do_syscall_64.entry_SYSCALL_64_after_hwframe
-0.2 0.57 ± 4% perf-profile.calltrace.cycles-pp.__switch_to_asm
-0.2 0.75 ± 5% perf-profile.calltrace.cycles-pp.__consume_stateless_skb.udp_recvmsg.inet_recvmsg.__sys_recvfrom.__x64_sys_recvfrom
-0.2 0.54 ± 4% perf-profile.calltrace.cycles-pp.tick_nohz_idle_exit.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64
-0.1 0.58 ± 9% perf-profile.calltrace.cycles-pp.nf_hook_slow.ip_output.ip_send_skb.udp_send_skb.udp_sendmsg
-0.1 0.82 ± 8% perf-profile.calltrace.cycles-pp.select_task_rq_fair.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock
-0.1 0.88 ± 5% perf-profile.calltrace.cycles-pp.ip_route_output_flow.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
-0.1 0.96 ± 5% perf-profile.calltrace.cycles-pp.ttwu_do_wakeup.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock
-0.1 0.62 ± 4% perf-profile.calltrace.cycles-pp.poll_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary
-0.1 0.92 ± 5% perf-profile.calltrace.cycles-pp.check_preempt_curr.ttwu_do_wakeup.try_to_wake_up.autoremove_wake_function.__wake_up_common
-0.1 0.59 ± 7% perf-profile.calltrace.cycles-pp.sk_filter_trim_cap.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver
-0.1 0.60 ± 8% perf-profile.calltrace.cycles-pp._raw_spin_lock_bh.__skb_recv_udp.udp_recvmsg.inet_recvmsg.__sys_recvfrom
-0.1 0.55 ± 6% perf-profile.calltrace.cycles-pp.security_sock_rcv_skb.sk_filter_trim_cap.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish
+1.9 28.61 ± 5% perf-profile.calltrace.cycles-pp.__x64_sys_sendto.do_syscall_64.entry_SYSCALL_64_after_hwframe
+1.9 28.45 ± 5% perf-profile.calltrace.cycles-pp.__sys_sendto.__x64_sys_sendto.do_syscall_64.entry_SYSCALL_64_after_hwframe
+2.1 27.80 ± 5% perf-profile.calltrace.cycles-pp.sock_sendmsg.__sys_sendto.__x64_sys_sendto.do_syscall_64.entry_SYSCALL_64_after_hwframe
+2.2 27.38 ± 5% perf-profile.calltrace.cycles-pp.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto.do_syscall_64
+3.3 21.89 ± 5% perf-profile.calltrace.cycles-pp.udp_send_skb.udp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
+3.4 21.68 ± 5% perf-profile.calltrace.cycles-pp.ip_send_skb.udp_send_skb.udp_sendmsg.sock_sendmsg.__sys_sendto
+3.4 21.32 ± 5% perf-profile.calltrace.cycles-pp.ip_output.ip_send_skb.udp_send_skb.udp_sendmsg.sock_sendmsg
+3.6 20.34 ± 5% perf-profile.calltrace.cycles-pp.ip_finish_output2.ip_output.ip_send_skb.udp_send_skb.udp_sendmsg
+4.0 18.32 ± 5% perf-profile.calltrace.cycles-pp.__local_bh_enable_ip.ip_finish_output2.ip_output.ip_send_skb.udp_send_skb
+4.0 18.22 ± 5% perf-profile.calltrace.cycles-pp.do_softirq.__local_bh_enable_ip.ip_finish_output2.ip_output.ip_send_skb
+4.0 18.01 ± 5% perf-profile.calltrace.cycles-pp.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip.ip_finish_output2
+4.0 18.07 ± 5% perf-profile.calltrace.cycles-pp.do_softirq_own_stack.do_softirq.__local_bh_enable_ip.ip_finish_output2.ip_output
+4.1 17.66 ± 5% perf-profile.calltrace.cycles-pp.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip
+4.1 17.42 ± 5% perf-profile.calltrace.cycles-pp.process_backlog.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq
+4.1 4.14 ± 5% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__percpu_counter_sum.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb
+4.2 16.77 ± 5% perf-profile.calltrace.cycles-pp.__netif_receive_skb_one_core.process_backlog.net_rx_action.__softirqentry_text_start.do_softirq_own_stack
+4.2 16.52 ± 5% perf-profile.calltrace.cycles-pp.ip_rcv.__netif_receive_skb_one_core.process_backlog.net_rx_action.__softirqentry_text_start
+4.3 16.14 ± 5% perf-profile.calltrace.cycles-pp.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core.process_backlog.net_rx_action
+4.3 16.00 ± 5% perf-profile.calltrace.cycles-pp.ip_local_deliver_finish.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core.process_backlog
+4.4 15.77 ± 6% perf-profile.calltrace.cycles-pp.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core
+4.4 4.44 ± 5% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__percpu_counter_sum.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb.udp_queue_rcv_skb
+4.4 15.34 ± 6% perf-profile.calltrace.cycles-pp.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv
+4.6 14.57 ± 6% perf-profile.calltrace.cycles-pp.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish.ip_local_deliver
+5.8 5.80 ± 5% perf-profile.calltrace.cycles-pp.__percpu_counter_sum.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv
+6.0 6.01 ± 5% perf-profile.calltrace.cycles-pp.__sk_mem_raise_allocated.__udp_enqueue_schedule_skb.udp_queue_rcv_skb.__udp4_lib_rcv.ip_local_deliver_finish
-2.5 11.22 ± 6% perf-profile.children.cycles-pp.__x64_sys_recvfrom
-2.4 11.06 ± 6% perf-profile.children.cycles-pp.__sys_recvfrom
-2.1 9.73 ± 6% perf-profile.children.cycles-pp.inet_recvmsg
-2.1 9.59 ± 6% perf-profile.children.cycles-pp.udp_recvmsg
-1.8 9.62 ± 6% perf-profile.children.cycles-pp.__sched_text_start
-1.6 7.47 ± 7% perf-profile.children.cycles-pp.__skb_recv_udp
-1.4 6.28 ± 7% perf-profile.children.cycles-pp.__skb_wait_for_more_packets
-1.3 7.70 ± 6% perf-profile.children.cycles-pp.__wake_up_common_lock
-1.3 7.99 ± 6% perf-profile.children.cycles-pp.sock_def_readable
-1.2 7.24 ± 7% perf-profile.children.cycles-pp.__wake_up_common
-1.2 6.90 ± 7% perf-profile.children.cycles-pp.try_to_wake_up
-1.1 7.00 ± 7% perf-profile.children.cycles-pp.autoremove_wake_function
-1.0 5.12 ± 7% perf-profile.children.cycles-pp.schedule
-0.9 4.62 ± 5% perf-profile.children.cycles-pp.schedule_idle
-0.8 3.96 ± 6% perf-profile.children.cycles-pp.ip_make_skb
-0.6 3.86 ± 7% perf-profile.children.cycles-pp.ttwu_do_activate
-0.6 3.63 ± 8% perf-profile.children.cycles-pp.enqueue_task_fair
-0.5 3.14 ± 8% perf-profile.children.cycles-pp.enqueue_entity
-0.5 2.18 ± 7% perf-profile.children.cycles-pp.dequeue_task_fair
-0.4 1.88 ± 7% perf-profile.children.cycles-pp.pick_next_task_fair
-0.4 1.95 ± 8% perf-profile.children.cycles-pp.dequeue_entity
-0.4 2.47 ± 6% perf-profile.children.cycles-pp.__ip_append_data
-0.4 1.91 ± 6% perf-profile.children.cycles-pp.__entry_SYSCALL_64_trampoline
-0.4 1.29 ± 4% perf-profile.children.cycles-pp.__ip_make_skb
-0.4 2.44 ± 7% perf-profile.children.cycles-pp.menu_select
-0.4 2.24 ± 5% perf-profile.children.cycles-pp.switch_mm_irqs_off
-0.3 2.11 ± 8% perf-profile.children.cycles-pp.update_load_avg
-0.3 1.74 ± 7% perf-profile.children.cycles-pp.__dev_queue_xmit
-0.3 1.40 ± 5% perf-profile.children.cycles-pp.syscall_return_via_sysret
-0.3 1.00 ± 4% perf-profile.children.cycles-pp.__ip_select_ident
-0.3 0.94 ± 4% perf-profile.children.cycles-pp.ip_idents_reserve
-0.3 1.06 ± 6% perf-profile.children.cycles-pp.set_next_entity
-0.3 1.84 ± 6% perf-profile.children.cycles-pp.sock_alloc_send_pskb
-0.3 1.52 ± 8% perf-profile.children.cycles-pp.tick_nohz_get_sleep_length
-0.3 1.23 ± 3% perf-profile.children.cycles-pp._raw_spin_lock
-0.2 1.03 ± 6% perf-profile.children.cycles-pp.load_new_mm_cr3
-0.2 1.37 ± 7% perf-profile.children.cycles-pp.dev_hard_start_xmit
-0.2 1.31 ± 7% perf-profile.children.cycles-pp.loopback_xmit
-0.2 1.29 ± 7% perf-profile.children.cycles-pp.alloc_skb_with_frags
-0.2 0.66 ± 8% perf-profile.children.cycles-pp.update_curr
-0.2 1.24 ± 7% perf-profile.children.cycles-pp.__alloc_skb
-0.2 0.55 ± 4% perf-profile.children.cycles-pp.tick_nohz_idle_exit
-0.2 0.73 ± 4% perf-profile.children.cycles-pp._copy_to_iter
-0.2 0.90 ± 6% perf-profile.children.cycles-pp.switch_mm
-0.2 0.94 ± 8% perf-profile.children.cycles-pp.get_next_timer_interrupt
-0.2 0.66 ± 4% perf-profile.children.cycles-pp.sock_wfree
-0.2 0.80 ± 6% perf-profile.children.cycles-pp.move_addr_to_user
-0.2 0.75 ± 5% perf-profile.children.cycles-pp.__consume_stateless_skb
-0.2 0.59 ± 4% perf-profile.children.cycles-pp.__switch_to_asm
-0.1 0.64 ± 6% perf-profile.children.cycles-pp.copy_user_generic_unrolled
-0.1 0.35 ± 5% perf-profile.children.cycles-pp.update_ts_time_stats
-0.1 0.53 ± 5% perf-profile.children.cycles-pp.__switch_to
-0.1 0.70 ± 8% perf-profile.children.cycles-pp.nf_hook_slow
-0.1 0.58 ± 9% perf-profile.children.cycles-pp.__next_timer_interrupt
-0.1 0.82 ± 7% perf-profile.children.cycles-pp.select_task_rq_fair
-0.1 0.55 ± 12% perf-profile.children.cycles-pp.update_cfs_group
-0.1 0.44 ± 7% perf-profile.children.cycles-pp.prepare_to_wait_exclusive
-0.1 0.31 ± 10% perf-profile.children.cycles-pp.__enqueue_entity
-0.1 0.56 ± 5% perf-profile.children.cycles-pp.__slab_free
-0.1 0.31 ± 7% perf-profile.children.cycles-pp.ip_generic_getfrag
-0.1 0.63 ± 4% perf-profile.children.cycles-pp.poll_idle
-0.1 0.28 ± 7% perf-profile.children.cycles-pp._copy_from_iter_full
-0.1 0.97 ± 5% perf-profile.children.cycles-pp.ttwu_do_wakeup
-0.1 0.94 ± 5% perf-profile.children.cycles-pp.check_preempt_curr
-0.1 0.44 ± 7% perf-profile.children.cycles-pp.update_rq_clock
-0.1 0.47 ± 9% perf-profile.children.cycles-pp.selinux_ip_postroute
-0.1 0.43 ± 4% perf-profile.children.cycles-pp.sock_def_write_space
-0.1 0.59 ± 7% perf-profile.children.cycles-pp.sk_filter_trim_cap
-0.1 0.48 ± 5% perf-profile.children.cycles-pp.sched_clock_cpu
-0.1 0.44 ± 4% perf-profile.children.cycles-pp.copyout
-0.1 0.55 ± 6% perf-profile.children.cycles-pp.security_sock_rcv_skb
-0.1 0.47 ± 4% perf-profile.children.cycles-pp.skb_set_owner_w
-0.1 0.60 ± 8% perf-profile.children.cycles-pp._raw_spin_lock_bh
-0.1 0.44 ± 4% perf-profile.children.cycles-pp.sched_clock
-0.1 0.30 ± 8% perf-profile.children.cycles-pp.__update_load_avg_se
-0.1 0.26 ± 6% perf-profile.children.cycles-pp.hrtimer_next_event_without
-0.1 0.34 ± 5% perf-profile.children.cycles-pp.tick_nohz_idle_enter
-0.1 0.53 ± 6% perf-profile.children.cycles-pp.selinux_socket_sock_rcv_skb
-0.1 0.30 ± 14% perf-profile.children.cycles-pp.___perf_sw_event
-0.1 0.42 ± 5% perf-profile.children.cycles-pp.native_sched_clock
-0.1 0.33 ± 4% perf-profile.children.cycles-pp.nr_iowait_cpu
-0.1 0.34 ± 8% perf-profile.children.cycles-pp.sockfd_lookup_light
-0.1 0.37 ± 11% perf-profile.children.cycles-pp.selinux_ip_postroute_compat
-0.1 0.37 ± 8% perf-profile.children.cycles-pp.native_write_msr
-0.1 0.48 ± 6% perf-profile.children.cycles-pp.finish_task_switch
-0.1 0.36 ± 9% perf-profile.children.cycles-pp.ksize
-0.1 0.23 ± 5% perf-profile.children.cycles-pp.__udp4_lib_lookup
-0.1 0.27 ± 8% perf-profile.children.cycles-pp.__fget_light
-0.1 0.25 ± 11% perf-profile.children.cycles-pp.kmem_cache_alloc_node
-0.1 0.14 ± 3% perf-profile.children.cycles-pp.copyin
-0.1 0.28 ± 8% perf-profile.children.cycles-pp.__might_fault
-0.1 0.22 ± 5% perf-profile.children.cycles-pp.pick_next_task_idle
-0.1 0.15 ± 2% perf-profile.children.cycles-pp.entry_SYSCALL_64_stage2
-0.1 0.43 ± 5% perf-profile.children.cycles-pp._raw_spin_unlock_irqrestore
-0.1 0.21 ± 7% perf-profile.children.cycles-pp.validate_xmit_skb
-0.1 0.04 ± 57% perf-profile.children.cycles-pp.kfree
-0.1 0.18 ± 6% perf-profile.children.cycles-pp.rb_erase_cached
-0.1 0.39 ± 5% perf-profile.children.cycles-pp.reweight_entity
-0.1 0.21 ± 7% perf-profile.children.cycles-pp.enqueue_to_backlog
-0.0 0.20 ± 5% perf-profile.children.cycles-pp.security_socket_sendmsg
-0.0 0.16 ± 6% perf-profile.children.cycles-pp.kmem_cache_free
-0.0 0.10 ± 10% perf-profile.children.cycles-pp.__hrtimer_next_event_base
-0.0 0.03 ±100% perf-profile.children.cycles-pp.rb_next
-0.0 0.20 ± 9% perf-profile.children.cycles-pp.___might_sleep
-0.0 0.14 perf-profile.children.cycles-pp.inet_sendmsg
-0.0 0.16 ± 6% perf-profile.children.cycles-pp.rcu_idle_exit
-0.0 0.15 ± 2% perf-profile.children.cycles-pp.copy_user_enhanced_fast_string
-0.0 0.13 ± 11% perf-profile.children.cycles-pp.put_prev_task_fair
-0.0 0.10 ± 10% perf-profile.children.cycles-pp.compute_score
-0.0 0.19 ± 10% perf-profile.children.cycles-pp.selinux_parse_skb
-0.0 0.25 ± 9% perf-profile.children.cycles-pp.ipv4_mtu
-0.0 0.12 ± 6% perf-profile.children.cycles-pp.__calc_delta
-0.0 0.21 ± 8% perf-profile.children.cycles-pp.move_addr_to_kernel
-0.0 0.19 ± 3% perf-profile.children.cycles-pp.sock_has_perm
-0.0 0.19 ± 9% perf-profile.children.cycles-pp.__list_del_entry_valid
-0.0 0.22 ± 4% perf-profile.children.cycles-pp.pick_next_entity
-0.0 0.18 ± 7% perf-profile.children.cycles-pp._copy_from_user
-0.0 0.16 ± 6% perf-profile.children.cycles-pp.__skb_try_recv_from_queue
-0.0 0.17 ± 5% perf-profile.children.cycles-pp.rcu_eqs_enter
-0.0 0.14 ± 5% perf-profile.children.cycles-pp.security_sk_classify_flow
-0.0 0.11 ± 7% perf-profile.children.cycles-pp.menu_reflect
-0.0 0.07 ± 7% perf-profile.children.cycles-pp.selinux_socket_sendmsg
-0.0 0.16 ± 7% perf-profile.children.cycles-pp._copy_to_user
-0.0 0.16 ± 8% perf-profile.children.cycles-pp.cpuidle_governor_latency_req
-0.0 0.15 ± 11% perf-profile.children.cycles-pp.__might_sleep
-0.0 0.14 perf-profile.children.cycles-pp.sock_recvmsg
-0.0 0.08 ± 6% perf-profile.children.cycles-pp.ipv4_pktinfo_prepare
-0.0 0.12 perf-profile.children.cycles-pp.security_socket_recvmsg
-0.0 0.11 ± 6% perf-profile.children.cycles-pp.__get_user_4
-0.0 0.11 ± 6% perf-profile.children.cycles-pp.selinux_sk_getsecid
-0.0 0.08 ± 5% perf-profile.children.cycles-pp.dst_release
-0.0 0.08 ± 5% perf-profile.children.cycles-pp.selinux_ipv4_postroute
-0.0 0.10 ± 11% perf-profile.children.cycles-pp.pm_qos_request
-0.0 0.09 ± 4% perf-profile.children.cycles-pp.udp_rcv
-0.0 0.07 ± 5% perf-profile.children.cycles-pp.ip_send_check
+0.0 0.07 ± 13% perf-profile.children.cycles-pp.selinux_netlbl_sock_rcv_skb
+0.0 0.08 ± 6% perf-profile.children.cycles-pp.rebalance_domains
+0.0 0.09 ± 7% perf-profile.children.cycles-pp.ip_rcv_finish_core
+0.0 0.12 ± 3% perf-profile.children.cycles-pp.ip_rcv_finish
+0.0 0.20 ± 8% perf-profile.children.cycles-pp.irq_exit
+0.1 0.16 ± 22% perf-profile.children.cycles-pp.clockevents_program_event
+0.1 0.06 ± 6% perf-profile.children.cycles-pp.netlbl_enabled
+0.1 0.57 ± 7% perf-profile.children.cycles-pp.find_next_bit
+0.2 0.17 ± 4% perf-profile.children.cycles-pp.percpu_counter_add_batch
+0.3 0.34 ± 5% perf-profile.children.cycles-pp.cpumask_next
+1.9 28.63 ± 5% perf-profile.children.cycles-pp.__x64_sys_sendto
+1.9 28.46 ± 5% perf-profile.children.cycles-pp.__sys_sendto
+2.1 27.81 ± 5% perf-profile.children.cycles-pp.sock_sendmsg
+2.2 27.40 ± 5% perf-profile.children.cycles-pp.udp_sendmsg
+3.3 21.90 ± 5% perf-profile.children.cycles-pp.udp_send_skb
+3.4 21.68 ± 5% perf-profile.children.cycles-pp.ip_send_skb
+3.4 21.33 ± 5% perf-profile.children.cycles-pp.ip_output
+3.6 20.36 ± 5% perf-profile.children.cycles-pp.ip_finish_output2
+4.0 18.36 ± 5% perf-profile.children.cycles-pp.__local_bh_enable_ip
+4.0 18.25 ± 5% perf-profile.children.cycles-pp.do_softirq
+4.0 18.08 ± 5% perf-profile.children.cycles-pp.do_softirq_own_stack
+4.0 18.16 ± 5% perf-profile.children.cycles-pp.__softirqentry_text_start
+4.1 17.67 ± 5% perf-profile.children.cycles-pp.net_rx_action
+4.1 17.43 ± 5% perf-profile.children.cycles-pp.process_backlog
+4.2 5.68 ± 5% perf-profile.children.cycles-pp._raw_spin_lock_irqsave
+4.2 4.19 ± 5% perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
+4.2 16.52 ± 5% perf-profile.children.cycles-pp.ip_rcv
+4.2 16.78 ± 5% perf-profile.children.cycles-pp.__netif_receive_skb_one_core
+4.3 16.15 ± 5% perf-profile.children.cycles-pp.ip_local_deliver
+4.3 16.00 ± 5% perf-profile.children.cycles-pp.ip_local_deliver_finish
+4.4 15.78 ± 6% perf-profile.children.cycles-pp.__udp4_lib_rcv
+4.4 15.35 ± 6% perf-profile.children.cycles-pp.udp_queue_rcv_skb
+4.6 14.58 ± 6% perf-profile.children.cycles-pp.__udp_enqueue_schedule_skb
+5.8 5.84 ± 5% perf-profile.children.cycles-pp.__percpu_counter_sum
+5.9 6.01 ± 5% perf-profile.children.cycles-pp.__sk_mem_raise_allocated
-0.4 1.85 ± 6% perf-profile.self.cycles-pp.__entry_SYSCALL_64_trampoline
-0.3 1.89 ± 5% perf-profile.self.cycles-pp.__sched_text_start
-0.3 1.39 ± 5% perf-profile.self.cycles-pp.syscall_return_via_sysret
-0.3 0.93 ± 4% perf-profile.self.cycles-pp.ip_idents_reserve
-0.3 1.16 ± 4% perf-profile.self.cycles-pp._raw_spin_lock
-0.2 1.03 ± 6% perf-profile.self.cycles-pp.load_new_mm_cr3
-0.2 0.82 ± 7% perf-profile.self.cycles-pp.do_idle
-0.2 0.29 ± 5% perf-profile.self.cycles-pp.__skb_wait_for_more_packets
-0.2 0.90 ± 6% perf-profile.self.cycles-pp.update_load_avg
-0.2 0.59 ± 4% perf-profile.self.cycles-pp.__switch_to_asm
-0.1 0.45 ± 6% perf-profile.self.cycles-pp.udp_sendmsg
-0.1 0.62 ± 5% perf-profile.self.cycles-pp.enqueue_entity
-0.1 1.21 ± 5% perf-profile.self.cycles-pp.switch_mm_irqs_off
-0.1 0.50 ± 5% perf-profile.self.cycles-pp.__switch_to
-0.1 0.62 ± 6% perf-profile.self.cycles-pp.copy_user_generic_unrolled
-0.1 0.54 ± 12% perf-profile.self.cycles-pp.update_cfs_group
-0.1 0.44 ± 8% perf-profile.self.cycles-pp.update_curr
-0.1 0.56 ± 5% perf-profile.self.cycles-pp.__slab_free
-0.1 0.30 ± 11% perf-profile.self.cycles-pp.__enqueue_entity
-0.1 0.54 ± 5% perf-profile.self.cycles-pp.set_next_entity
-0.1 0.23 ± 3% perf-profile.self.cycles-pp.ip_output
-0.1 0.49 ± 4% perf-profile.self.cycles-pp.enqueue_task_fair
-0.1 0.42 ± 9% perf-profile.self.cycles-pp.pick_next_task_fair
-0.1 0.42 ± 7% perf-profile.self.cycles-pp.do_syscall_64
-0.1 0.55 ± 4% perf-profile.self.cycles-pp.poll_idle
-0.1 0.45 ± 5% perf-profile.self.cycles-pp.move_addr_to_user
-0.1 0.54 ± 6% perf-profile.self.cycles-pp.select_task_rq_fair
-0.1 0.31 ± 5% perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
-0.1 0.47 ± 4% perf-profile.self.cycles-pp.skb_set_owner_w
-0.1 0.43 ± 5% perf-profile.self.cycles-pp.sock_def_write_space
-0.1 0.11 ± 7% perf-profile.self.cycles-pp.check_preempt_curr
-0.1 0.60 ± 8% perf-profile.self.cycles-pp._raw_spin_lock_bh
-0.1 0.45 ± 5% perf-profile.self.cycles-pp.udp_recvmsg
-0.1 0.30 ± 8% perf-profile.self.cycles-pp.__update_load_avg_se
-0.1 0.23 ± 8% perf-profile.self.cycles-pp.__skb_recv_udp
-0.1 0.43 ± 10% perf-profile.self.cycles-pp.dequeue_entity
-0.1 0.18 ± 6% perf-profile.self.cycles-pp.__ip_make_skb
-0.1 0.41 ± 4% perf-profile.self.cycles-pp.native_sched_clock
-0.1 0.22 ± 5% perf-profile.self.cycles-pp._copy_to_iter
-0.1 0.37 ± 7% perf-profile.self.cycles-pp.native_write_msr
-0.1 0.37 ± 5% perf-profile.self.cycles-pp._raw_spin_unlock_irqrestore
-0.1 0.33 ± 4% perf-profile.self.cycles-pp.nr_iowait_cpu
-0.1 0.18 ± 8% perf-profile.self.cycles-pp.__wake_up_common
-0.1 0.35 ± 9% perf-profile.self.cycles-pp.ksize
-0.1 0.26 ± 9% perf-profile.self.cycles-pp.__next_timer_interrupt
-0.1 0.21 ± 5% perf-profile.self.cycles-pp.sock_wfree
-0.1 0.26 ± 8% perf-profile.self.cycles-pp.__fget_light
-0.1 0.25 ± 5% perf-profile.self.cycles-pp.update_rq_clock
-0.1 0.43 ± 6% perf-profile.self.cycles-pp.__udp_enqueue_schedule_skb
-0.1 0.15 ± 5% perf-profile.self.cycles-pp.__sys_recvfrom
-0.1 0.24 ± 6% perf-profile.self.cycles-pp.try_to_wake_up
-0.1 0.04 ± 57% perf-profile.self.cycles-pp.kfree
-0.1 0.15 ± 2% perf-profile.self.cycles-pp.entry_SYSCALL_64_stage2
-0.0 0.39 ± 6% perf-profile.self.cycles-pp.reweight_entity
-0.0 0.31 ± 6% perf-profile.self.cycles-pp.__softirqentry_text_start
-0.0 0.12 ± 8% perf-profile.self.cycles-pp.__local_bh_enable_ip
-0.0 0.21 ± 7% perf-profile.self.cycles-pp.net_rx_action
-0.0 0.04 ± 57% perf-profile.self.cycles-pp.tick_nohz_idle_exit
-0.0 0.21 ± 7% perf-profile.self.cycles-pp.__sys_sendto
-0.0 0.16 ± 5% perf-profile.self.cycles-pp.__x64_sys_recvfrom
-0.0 0.08 ± 8% perf-profile.self.cycles-pp.udp_queue_rcv_skb
-0.0 0.22 ± 7% perf-profile.self.cycles-pp.loopback_xmit
-0.0 0.14 ± 5% perf-profile.self.cycles-pp.inet_recvmsg
-0.0 0.10 ± 8% perf-profile.self.cycles-pp.compute_score
-0.0 0.15 ± 8% perf-profile.self.cycles-pp.selinux_socket_sock_rcv_skb
-0.0 0.11 ± 7% perf-profile.self.cycles-pp.ip_local_deliver_finish
-0.0 0.16 ± 7% perf-profile.self.cycles-pp.rb_erase_cached
-0.0 0.18 ± 8% perf-profile.self.cycles-pp.__alloc_skb
-0.0 0.14 ± 9% perf-profile.self.cycles-pp.kmem_cache_alloc_node
-0.0 0.16 ± 8% perf-profile.self.cycles-pp.rcu_eqs_enter
-0.0 0.14 ± 3% perf-profile.self.cycles-pp.copy_user_enhanced_fast_string
-0.0 0.13 perf-profile.self.cycles-pp.inet_sendmsg
-0.0 0.26 ± 4% perf-profile.self.cycles-pp.process_backlog
-0.0 0.19 ± 2% perf-profile.self.cycles-pp.pick_next_entity
-0.0 0.11 ± 9% perf-profile.self.cycles-pp.selinux_ip_postroute_compat
-0.0 0.12 ± 6% perf-profile.self.cycles-pp.__calc_delta
-0.0 0.08 ± 8% perf-profile.self.cycles-pp.validate_xmit_skb
-0.0 0.05 ± 9% perf-profile.self.cycles-pp._copy_from_iter_full
-0.0 0.05 ± 9% perf-profile.self.cycles-pp.selinux_socket_sendmsg
-0.0 0.19 ± 3% perf-profile.self.cycles-pp.sock_has_perm
-0.0 0.12 ± 4% perf-profile.self.cycles-pp.__udp4_lib_lookup
-0.0 0.12 ± 10% perf-profile.self.cycles-pp.__might_sleep
-0.0 0.06 ± 6% perf-profile.self.cycles-pp.nf_hook_slow
-0.0 0.10 ± 8% perf-profile.self.cycles-pp.prepare_to_wait_exclusive
-0.0 0.12 ± 5% perf-profile.self.cycles-pp.enqueue_to_backlog
-0.0 0.06 ± 15% perf-profile.self.cycles-pp.ip_make_skb
-0.0 0.11 ± 3% perf-profile.self.cycles-pp.ip_rcv
-0.0 0.11 ± 4% perf-profile.self.cycles-pp.selinux_ip_postroute
-0.0 0.08 ± 6% perf-profile.self.cycles-pp.sock_sendmsg
-0.0 0.08 ± 10% perf-profile.self.cycles-pp.__might_fault
-0.0 0.07 ± 13% perf-profile.self.cycles-pp.put_prev_task_fair
-0.0 0.10 ± 5% perf-profile.self.cycles-pp.rcu_idle_exit
-0.0 0.10 ± 7% perf-profile.self.cycles-pp.menu_reflect
-0.0 0.04 ± 57% perf-profile.self.cycles-pp.__ip_local_out
-0.0 0.11 ± 7% perf-profile.self.cycles-pp.selinux_sock_rcv_skb_compat
-0.0 0.06 ± 14% perf-profile.self.cycles-pp.__skb_try_recv_from_queue
-0.0 0.07 ± 5% perf-profile.self.cycles-pp.selinux_ipv4_postroute
-0.0 0.08 ± 6% perf-profile.self.cycles-pp.dst_release
-0.0 0.07 ± 7% perf-profile.self.cycles-pp.schedule
-0.0 0.11 ± 6% perf-profile.self.cycles-pp.__get_user_4
-0.0 0.09 ± 7% perf-profile.self.cycles-pp.netif_skb_features
-0.0 0.09 ± 7% perf-profile.self.cycles-pp.__hrtimer_next_event_base
-0.0 0.17 ± 5% perf-profile.self.cycles-pp.__x64_sys_sendto
-0.0 0.10 ± 5% perf-profile.self.cycles-pp.switch_mm
-0.0 0.10 ± 7% perf-profile.self.cycles-pp.selinux_sk_getsecid
-0.0 0.08 ± 6% perf-profile.self.cycles-pp.udp_rcv
-0.0 0.05 ± 9% perf-profile.self.cycles-pp.update_min_vruntime
-0.0 0.07 perf-profile.self.cycles-pp.ip_send_check
-0.0 0.06 ± 6% perf-profile.self.cycles-pp.sockfd_lookup_light
+0.0 0.09 ± 7% perf-profile.self.cycles-pp.ip_rcv_finish_core
+0.1 0.06 ± 9% perf-profile.self.cycles-pp.netlbl_enabled
+0.1 0.06 ± 15% perf-profile.self.cycles-pp.hrtimer_get_next_event
+0.1 0.29 ± 3% perf-profile.self.cycles-pp.sock_def_readable
+0.1 0.11 ± 7% perf-profile.self.cycles-pp.cpumask_next
+0.1 0.50 ± 7% perf-profile.self.cycles-pp.find_next_bit
+0.2 0.17 ± 4% perf-profile.self.cycles-pp.percpu_counter_add_batch
+1.1 1.07 ± 5% perf-profile.self.cycles-pp.__percpu_counter_sum
+4.2 4.19 ± 5% perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
have been estimated based on internal Intel analysis and are provided
purposes only. Any difference in system hardware or software
configuration may affect actual performance.

Attachments:

(No filename) (140.76 kB)
config-4.19.0-rc2-00128-gb99259a (163.74 kB)
job-script (7.27 kB)
job.yaml (4.89 kB)
reproduce (1.29 kB)
Download all attachments