LinuxLists.cc - split up lockdep and syscall related functionailty in generic entry code

2020-12-01 08:41:37

Subject: split up lockdep and syscall related functionailty in generic entry code

i'm currently working on converting s390 to use the generic entry
functionality. So far things are straigt-forward, there's only one
slight problem. There is a syscall_enter_from_user_mode() which sets
lockdep state and other initial stuff + does the entry work at the same
time. This is a problem on s390 because the way we restart syscalls isn't
as easy as on x86.

My understanding on x86 is that syscalls are restarted there by just rewinding
the program counter and return to user space, so the instruction causing
the syscall gets executed again.

On s390 this doesn't work, because the syscall number might be hard coded
into the 'svc' instruction, so when the syscall number has to be changed we
would repeat the wrong (old) syscall.

So we would need functions that only do the stuff that is required when switching
from user space to kernel and back, and functions which do the system call tracing
and work which might be called repeatedly.

With the attached patch, the s390 code now looks like this:

(i removed some s390 specific stuff here to make the function easier
to read)

__do_syscall is the function which gets called by low level entry.S code:

void noinstr __do_syscall(struct pt_regs *regs)
{
enter_from_user_mode(regs); /* sets lockdep state, and other initial stuff */

/*
* functions that need to run with irqs disabled,
* but lockdep state and other stuff set up
*/
memcpy(&regs->gprs[8], S390_lowcore.save_area_sync, 8 * sizeof(unsigned long));
memcpy(&regs->int_code, &S390_lowcore.svc_ilc, sizeof(regs->int_code));
regs->psw = S390_lowcore.svc_old_psw;

update_timer_sys();

local_irq_enable();

regs->orig_gpr2 = regs->gprs[2];

do {
regs->flags = _PIF_SYSCALL;
do_syscall(regs);
} while (test_pt_regs_flag(regs, PIF_SYSCALL_RESTART));

exit_to_user_mode();
}

__do_syscall calls do_syscall which does all the syscall work, and this might
be called more than once if PIF_SYSCALL_RESTART is set:

void do_syscall(struct pt_regs *regs)
{
unsigned long nr = regs->int_code & 0xffff;

nr = syscall_enter_from_user_mode_work(regs, nr);
regs->gprs[2] = -ENOSYS;
if (likely(nr < NR_syscalls)) {
regs->gprs[2] = current->thread.sys_call_table[nr](
regs->orig_gpr2, regs->gprs[3],
regs->gprs[4], regs->gprs[5],
regs->gprs[6], regs->gprs[7]);
}
syscall_exit_to_user_mode1(regs);
}

What do you think about the attach patch? I'm also open for a proper name
for syscall_exit_to_user_mode1() ;-)

2020-12-01 08:42:10

by Sven Schnelle

[permalink] [raw]

Subject: [PATCH] entry: split lockdep and syscall work functions

On s390, we can not call one function which sets lockdep
state and do the syscall work at the same time. There add
make enter_from_user_mode() and exit_to_user_mode() public, and
add syscall_exit_to_user_mode1() which does the same as
syscall_exit_to_user_mode() but skips the final exit_to_user_mode().

Signed-off-by: Sven Schnelle <[email protected]>
---
include/linux/entry-common.h | 4 +++-
kernel/entry/common.c | 35 +++++++++++++++++++++++++++--------
2 files changed, 30 insertions(+), 9 deletions(-)

diff --git a/include/linux/entry-common.h b/include/linux/entry-common.h
index 474f29638d2c..496c9a47eab4 100644
--- a/include/linux/entry-common.h
+++ b/include/linux/entry-common.h
@@ -124,7 +124,7 @@ static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs
* to be done between establishing state and handling user mode entry work.
*/
void syscall_enter_from_user_mode_prepare(struct pt_regs *regs);
-
+void enter_from_user_mode(struct pt_regs *regs);
/**
* syscall_enter_from_user_mode_work - Check and handle work before invoking
* a syscall
@@ -311,6 +311,8 @@ static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
* arch_exit_to_user_mode() to handle e.g. speculation mitigations
*/
void syscall_exit_to_user_mode(struct pt_regs *regs);
+void syscall_exit_to_user_mode1(struct pt_regs *regs);
+void exit_to_user_mode(void);

/**
* irqentry_enter_from_user_mode - Establish state before invoking the irq handler
diff --git a/kernel/entry/common.c b/kernel/entry/common.c
index e9e2df3f3f9e..3ad462ebfa15 100644
--- a/kernel/entry/common.c
+++ b/kernel/entry/common.c
@@ -18,7 +18,7 @@
* 2) Invoke context tracking if enabled to reactivate RCU
* 3) Trace interrupts off state
*/
-static __always_inline void enter_from_user_mode(struct pt_regs *regs)
+static __always_inline void __enter_from_user_mode(struct pt_regs *regs)
{
arch_check_user_regs(regs);
lockdep_hardirqs_off(CALLER_ADDR0);
@@ -31,6 +31,11 @@ static __always_inline void enter_from_user_mode(struct pt_regs *regs)
instrumentation_end();
}

+void noinstr enter_from_user_mode(struct pt_regs *regs)
+{
+ __enter_from_user_mode(regs);
+}
+
static inline void syscall_enter_audit(struct pt_regs *regs, long syscall)
{
if (unlikely(audit_context())) {
@@ -92,7 +97,7 @@ noinstr long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall)
{
long ret;

- enter_from_user_mode(regs);
+ __enter_from_user_mode(regs);

instrumentation_begin();
local_irq_enable();
@@ -104,14 +109,14 @@ noinstr long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall)

noinstr void syscall_enter_from_user_mode_prepare(struct pt_regs *regs)
{
- enter_from_user_mode(regs);
+ __enter_from_user_mode(regs);
instrumentation_begin();
local_irq_enable();
instrumentation_end();
}

/**
- * exit_to_user_mode - Fixup state when exiting to user mode
+ * __exit_to_user_mode - Fixup state when exiting to user mode
*
* Syscall/interupt exit enables interrupts, but the kernel state is
* interrupts disabled when this is invoked. Also tell RCU about it.
@@ -122,7 +127,7 @@ noinstr void syscall_enter_from_user_mode_prepare(struct pt_regs *regs)
* mitigations, etc.
* 4) Tell lockdep that interrupts are enabled
*/
-static __always_inline void exit_to_user_mode(void)
+static __always_inline void __exit_to_user_mode(void)
{
instrumentation_begin();
trace_hardirqs_on_prepare();
@@ -134,6 +139,11 @@ static __always_inline void exit_to_user_mode(void)
lockdep_hardirqs_on(CALLER_ADDR0);
}

+void noinstr exit_to_user_mode(void)
+{
+ __exit_to_user_mode();
+}
+
/* Workaround to allow gradual conversion of architecture code */
void __weak arch_do_signal(struct pt_regs *regs) { }

@@ -265,12 +275,21 @@ __visible noinstr void syscall_exit_to_user_mode(struct pt_regs *regs)
local_irq_disable_exit_to_user();
exit_to_user_mode_prepare(regs);
instrumentation_end();
- exit_to_user_mode();
+ __exit_to_user_mode();
+}
+
+__visible noinstr void syscall_exit_to_user_mode1(struct pt_regs *regs)
+{
+ instrumentation_begin();
+ syscall_exit_to_user_mode_prepare(regs);
+ local_irq_disable_exit_to_user();
+ exit_to_user_mode_prepare(regs);
+ instrumentation_end();
}

noinstr void irqentry_enter_from_user_mode(struct pt_regs *regs)
{
- enter_from_user_mode(regs);
+ __enter_from_user_mode(regs);
}

noinstr void irqentry_exit_to_user_mode(struct pt_regs *regs)
@@ -278,7 +297,7 @@ noinstr void irqentry_exit_to_user_mode(struct pt_regs *regs)
instrumentation_begin();
exit_to_user_mode_prepare(regs);
instrumentation_end();
- exit_to_user_mode();
+ __exit_to_user_mode();
}

noinstr irqentry_state_t irqentry_enter(struct pt_regs *regs)
--
2.17.1

2020-12-01 11:13:09

by Thomas Gleixner

[permalink] [raw]

Subject: Re: [PATCH] entry: split lockdep and syscall work functions

Sven,

On Tue, Dec 01 2020 at 09:35, Sven Schnelle wrote:
> On s390, we can not call one function which sets lockdep
> state and do the syscall work at the same time. There add
> make enter_from_user_mode() and exit_to_user_mode() public, and
> add syscall_exit_to_user_mode1() which does the same as
> syscall_exit_to_user_mode() but skips the final exit_to_user_mode().

the explanation in the "cover letter" made at least sense, but the above
is unparseable word salad.

> Signed-off-by: Sven Schnelle <[email protected]>
> ---
> include/linux/entry-common.h | 4 +++-
> kernel/entry/common.c | 35 +++++++++++++++++++++++++++--------
> 2 files changed, 30 insertions(+), 9 deletions(-)
>
> diff --git a/include/linux/entry-common.h b/include/linux/entry-common.h
> index 474f29638d2c..496c9a47eab4 100644
> --- a/include/linux/entry-common.h
> +++ b/include/linux/entry-common.h
> @@ -124,7 +124,7 @@ static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs
> * to be done between establishing state and handling user mode entry work.
> */
> void syscall_enter_from_user_mode_prepare(struct pt_regs *regs);
> -
> +void enter_from_user_mode(struct pt_regs *regs);

You might have noticed, that all of these function prototypes have
proper kernel documentation. So just glueing this on to the previous
prototype does not cut it. enter_from/exit_to_user_mode() want to go
together into a seperate section.

> /**
> * syscall_enter_from_user_mode_work - Check and handle work before invoking
> * a syscall
> @@ -311,6 +311,8 @@ static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
> * arch_exit_to_user_mode() to handle e.g. speculation mitigations
> */
> void syscall_exit_to_user_mode(struct pt_regs *regs);
> +void syscall_exit_to_user_mode1(struct pt_regs *regs);

Same here and as you mentioned ...mode1() is a pretty horrible name.

syscall_exit_to_user_mode_work() perhaps?

> +void exit_to_user_mode(void);
>
> /**
> * irqentry_enter_from_user_mode - Establish state before invoking the irq handler
> diff --git a/kernel/entry/common.c b/kernel/entry/common.c
> index e9e2df3f3f9e..3ad462ebfa15 100644
> --- a/kernel/entry/common.c
> +++ b/kernel/entry/common.c
> @@ -18,7 +18,7 @@
> * 2) Invoke context tracking if enabled to reactivate RCU
> * 3) Trace interrupts off state
> */
> -static __always_inline void enter_from_user_mode(struct pt_regs *regs)
> +static __always_inline void __enter_from_user_mode(struct pt_regs
> *regs)

Can you please split the renaming into a seperate preparatory patch?

> +__visible noinstr void syscall_exit_to_user_mode1(struct pt_regs *regs)

What's the point of marking this function noinstr? Everything it does is
instrumentable.

Thanks,

tglx

2020-12-17 09:37:05

by kernel test robot

[permalink] [raw]

Subject: [entry] a358d5636c: unixbench.score -1.9% regression

Greeting,

FYI, we noticed a

commit: a358d5636cfcb8b278c url: in testcase: will-it-scale
on test machine: 192 with following parameters:

nr_task: 16
mode: process
test: futex3
cpufreq_governor: performance
ucode: 0x5003003

test-description: test-url: https://github.com/0day-ci/linux/commits/Sven-Schnelle/entry-split-lockdep-and-syscall-work-functions/20201201-163931
.kernel.org/cgit/linux/kernel/git/torvalds/linux.git">https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git 09162bc32c880a791c6c0668ce0745cf7958f576
threads Intel(R) Xeon(R) Platinum 9242 CPU @ 2.30GHz with 192G memory
Will It Scale takes a testcase and runs it from 1 through to n parallel copies to see if the testcase will scale. It builds both a process and threads based test in order to see any differences between the two.
/github.com/antonblanchard/will-it-scale">https://github.com/antonblanchard/will-it-scale
the commit also has significant impact on the following tests:
kindly add following tag
test robot <[email protected]>
----------------------------------------------------------------------->
href="https://github.com/intel/lkp-tests.git">https://github.com/intel/lkp-tests.git
# job file is attached in this email
==============================================================
config/mode/nr_task/rootfs/tbox_group/test/testcase/ucode:
rhel-8.3/process/16/debian-10.4-x86_64-20200603.cgz/lkp-csl-2ap2/futex3/will-it-scale/0x5003003
split lockdep and syscall work functions")
fcb8b278c766ac261
-----------------
fail:runs
|
9:9 stderr.Events_disabled
9:9 stderr.Events_enabled
9:9 stderr.[perf_record:Captured_and_wrote#MB/tmp/lkp/perf-sched.data(#samples)]
9:9 stderr.[perf_record:Woken_up#times_to_write_data]
9:9 stderr.has_stderr
%stddev
\
4947 ? 4% boot-time.idle
171377 ? 4% cpuidle.POLL.time
84528 ? 6% numa-numastat.node1.other_node
0.00 softirqs.CPU14.NET_RX
12085 slabinfo.kmalloc-2k.active_objs
12252 slabinfo.kmalloc-2k.num_objs
12431 ? 5% numa-vmstat.node3.nr_active_anon
88374 ? 5% numa-vmstat.node3.nr_file_pages
13240 ? 6% numa-vmstat.node3.nr_shmem
12431 ? 5% numa-vmstat.node3.nr_zone_active_anon
49674 ? 5% numa-meminfo.node3.Active
49674 ? 5% numa-meminfo.node3.Active(anon)
353508 ? 5% numa-meminfo.node3.FilePages
927015 ? 13% numa-meminfo.node3.MemUsed
52973 ? 6% numa-meminfo.node3.Shmem
53758 ? 3% meminfo.Active
53758 ? 3% meminfo.Active(anon)
+9.8% 516343 meminfo.Committed_AS
+15.1% 3220140 meminfo.Memused
65771 ? 3% meminfo.Shmem
14898 meminfo.max_used_kB
13426 ? 3% proc-vmstat.nr_active_anon
69896 proc-vmstat.nr_anon_pages
+5.6% 310056 proc-vmstat.nr_file_pages
72758 proc-vmstat.nr_inactive_anon
16436 ? 3% proc-vmstat.nr_shmem
71683 proc-vmstat.nr_slab_unreclaimable
+2.8% 293640 proc-vmstat.nr_unevictable
13426 ? 3% proc-vmstat.nr_zone_active_anon
72758 proc-vmstat.nr_zone_inactive_anon
+2.8% 293640 proc-vmstat.nr_zone_unevictable
+25.2% 1238643 proc-vmstat.numa_hit
+9.2% 978823 proc-vmstat.numa_local
259820 proc-vmstat.numa_other
17697 ? 4% proc-vmstat.pgactivate
+24.6% 1321886 proc-vmstat.pgalloc_normal
+2.7% 1192221 proc-vmstat.pgfault
+20.8% 1422376 proc-vmstat.pgfree
+25.9% 1.142e+10 perf-stat.i.branch-instructions
-20.5% 4349395 ? 6% perf-stat.i.cache-misses
0.78 perf-stat.i.cpi
+1.8% 201.74 perf-stat.i.cpu-migrations
13314 ? 5% perf-stat.i.cycles-between-cache-misses
+24.5% 1.853e+10 perf-stat.i.dTLB-loads
+25.2% 1.408e+10 perf-stat.i.dTLB-stores
1.29 perf-stat.i.ipc
1.44 ? 12% perf-stat.i.major-faults
229.60 perf-stat.i.metric.M/sec
3822 perf-stat.i.minor-faults
3823 perf-stat.i.page-faults
0.78 perf-stat.overall.cpi
13066 ? 6% perf-stat.overall.cycles-between-cache-misses
1.29 perf-stat.overall.ipc
+25.9% 1.138e+10 perf-stat.ps.branch-instructions
-20.5% 4334998 ? 6% perf-stat.ps.cache-misses
+1.7% 201.11 perf-stat.ps.cpu-migrations
+24.5% 1.847e+10 perf-stat.ps.dTLB-loads
+25.2% 1.403e+10 perf-stat.ps.dTLB-stores
1.43 ? 12% perf-stat.ps.major-faults
3809 perf-stat.ps.minor-faults
3810 perf-stat.ps.page-faults
0.00 sched_debug.cfs_rq:/.exec_clock.avg
0.00 sched_debug.cfs_rq:/.exec_clock.max
0.00 sched_debug.cfs_rq:/.exec_clock.min
0.00 sched_debug.cfs_rq:/.exec_clock.stddev
0.00 sched_debug.cfs_rq:/.nr_spread_over.avg
0.00 sched_debug.cfs_rq:/.nr_spread_over.max
0.00 sched_debug.cfs_rq:/.nr_spread_over.stddev
0.00 sched_debug.cpu.sched_count.avg
0.00 sched_debug.cpu.sched_count.max
0.00 sched_debug.cpu.sched_count.min
0.00 sched_debug.cpu.sched_count.stddev
0.00 sched_debug.cpu.sched_goidle.avg
0.00 sched_debug.cpu.sched_goidle.max
0.00 sched_debug.cpu.sched_goidle.min
0.00 sched_debug.cpu.sched_goidle.stddev
0.00 sched_debug.cpu.ttwu_count.avg
0.00 sched_debug.cpu.ttwu_count.max
0.00 sched_debug.cpu.ttwu_count.min
0.00 sched_debug.cpu.ttwu_count.stddev
0.00 sched_debug.cpu.ttwu_local.avg
0.00 sched_debug.cpu.ttwu_local.max
0.00 sched_debug.cpu.ttwu_local.min
0.00 sched_debug.cpu.ttwu_local.stddev
0.62 ? 7% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode_prepare.syscall_exit_to_user_mode.entry_SYSCALL_64_after_hwframe.syscall
0.06 ? 18% perf-profile.children.cycles-pp.unwind_next_frame
0.09 ? 13% perf-profile.children.cycles-pp.perf_callchain_kernel
0.10 ? 13% perf-profile.children.cycles-pp.get_perf_callchain
0.10 ? 13% perf-profile.children.cycles-pp.perf_callchain
0.11 ? 12% perf-profile.children.cycles-pp.perf_prepare_sample
0.11 ? 23% perf-profile.children.cycles-pp.process_simple
0.11 ? 25% perf-profile.children.cycles-pp.perf_session__process_events
0.11 ? 25% perf-profile.children.cycles-pp.record__finish_output
0.33 ? 8% perf-profile.children.cycles-pp.syscall@plt
0.12 ? 13% perf-profile.children.cycles-pp.perf_trace_sched_stat_runtime
0.12 ? 12% perf-profile.children.cycles-pp.perf_event_output_forward
0.13 ? 12% perf-profile.children.cycles-pp.perf_swevent_overflow
0.13 ? 12% perf-profile.children.cycles-pp.__perf_event_overflow
0.13 ? 14% perf-profile.children.cycles-pp.update_curr
0.13 ? 12% perf-profile.children.cycles-pp.perf_tp_event
0.14 ? 24% perf-profile.children.cycles-pp.cmd_sched
0.14 ? 24% perf-profile.children.cycles-pp.cmd_record
0.36 ? 17% perf-profile.children.cycles-pp.scheduler_tick
0.15 ? 12% perf-profile.children.cycles-pp.task_tick_fair
0.16 ? 21% perf-profile.children.cycles-pp.__libc_start_main
0.16 ? 21% perf-profile.children.cycles-pp.main
0.16 ? 21% perf-profile.children.cycles-pp.run_builtin
0.62 ? 7% perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
1.22 ? 8% perf-profile.self.cycles-pp.syscall_exit_to_user_mode
0.62 ? 7% perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
0.00 interrupts.161:PCI-MSI.524289-edge.eth0-TxRx-0
0.00 interrupts.162:PCI-MSI.524290-edge.eth0-TxRx-1
0.00 interrupts.163:PCI-MSI.524291-edge.eth0-TxRx-2
0.00 interrupts.164:PCI-MSI.524292-edge.eth0-TxRx-3
190.56 ? 68% interrupts.CPU10.RES:Rescheduling_interrupts
0.00 interrupts.CPU11.161:PCI-MSI.524289-edge.eth0-TxRx-0
1544 ? 20% interrupts.CPU115.CAL:Function_call_interrupts
1551 ? 20% interrupts.CPU116.CAL:Function_call_interrupts
0.00 interrupts.CPU12.162:PCI-MSI.524290-edge.eth0-TxRx-1
1588 ? 18% interrupts.CPU123.CAL:Function_call_interrupts
1587 ? 19% interrupts.CPU124.CAL:Function_call_interrupts
1593 ? 16% interrupts.CPU125.CAL:Function_call_interrupts
1552 ? 19% interrupts.CPU126.CAL:Function_call_interrupts
0.00 interrupts.CPU13.163:PCI-MSI.524291-edge.eth0-TxRx-2
0.00 interrupts.CPU14.164:PCI-MSI.524292-edge.eth0-TxRx-3
1634 ? 13% interrupts.CPU144.CAL:Function_call_interrupts
1572 ? 19% interrupts.CPU146.CAL:Function_call_interrupts
1545 ? 20% interrupts.CPU151.CAL:Function_call_interrupts
1558 ? 20% interrupts.CPU165.CAL:Function_call_interrupts
511.89 ? 24% interrupts.CPU191.NMI:Non-maskable_interrupts
511.89 ? 24% interrupts.CPU191.PMI:Performance_monitoring_interrupts
1556 ? 19% interrupts.CPU41.CAL:Function_call_interrupts
1544 ? 20% interrupts.CPU45.CAL:Function_call_interrupts
1606 ? 20% interrupts.CPU49.CAL:Function_call_interrupts
1560 ? 20% interrupts.CPU75.CAL:Function_call_interrupts
327.00 ? 22% interrupts.CPU95.NMI:Non-maskable_interrupts
327.00 ? 22% interrupts.CPU95.PMI:Performance_monitoring_interrupts
4053 ? 17% interrupts.RES:Rescheduling_interrupts

will-it-scale.per_process_ops

-------------------------------------------------+
+..+ +.. |
|
+ +..+.. |
+ +.. +.. +..|
+ .. .+.. .. +.. .. |
+ +. +.+ +.+ |
|
O O O O O |
O O O O O O O O O O O |
O |
|
|
|
-------------------------------------------------+

************************************************************************
Intel(R) Xeon(R) CPU @ 2.30GHz with 128G memory
==============================================================
config/nr_task/rootfs/runtime/tbox_group/test/testcase/ucode:
rhel-8.3/1/debian-10.4-x86_64-20200603.cgz/300s/lkp-csl-2sp4/syscall/unixbench/0x4003003
split lockdep and syscall work functions")
fcb8b278c766ac261
-----------------
fail:runs
|
4:4 stderr.Events_disabled
4:4 stderr.Events_enabled
4:4 stderr.[perf_record:Captured_and_wrote#MB/tmp/lkp/perf-sched.data(#samples)]
4:4 stderr.[perf_record:Woken_up#times_to_write_data]
4:4 stderr.has_stderr
0:4 perf-profile.children.cycles-pp.error_entry
0:4 perf-profile.self.cycles-pp.error_entry
%stddev
\
1478 unixbench.score
50172 unixbench.time.minor_page_faults
93.76 unixbench.time.user_time
-1.8% 8.654e+08 unixbench.workload
22.13 boot-time.boot
20310 ? 7% softirqs.CPU62.RCU
7448 slabinfo.kmalloc-2k.active_objs
7567 slabinfo.kmalloc-2k.num_objs
1071 ? 5% slabinfo.mnt_cache.active_objs
1071 ? 5% slabinfo.mnt_cache.num_objs
8895 ? 7% numa-meminfo.node1.Active
8839 ? 7% numa-meminfo.node1.Active(anon)
503916 ? 9% numa-meminfo.node1.FilePages
15358 ? 20% numa-meminfo.node1.Shmem
488571 ? 9% numa-meminfo.node1.Unevictable
11943 ? 5% meminfo.Active
11193 ? 5% meminfo.Active(anon)
748.75 ? 74% meminfo.Active(file)
+12.7% 2350606 meminfo.Memused
20830 ? 3% meminfo.Shmem
7199 ? 2% meminfo.max_used_kB
2211 ? 8% numa-vmstat.node1.nr_active_anon
125979 ? 9% numa-vmstat.node1.nr_file_pages
3840 ? 20% numa-vmstat.node1.nr_shmem
122142 ? 9% numa-vmstat.node1.nr_unevictable
2211 ? 8% numa-vmstat.node1.nr_zone_active_anon
122142 ? 9% numa-vmstat.node1.nr_zone_unevictable
-26.4% 1303772 ? 23% perf-stat.i.cache-misses
5402 ? 13% perf-stat.i.cycles-between-cache-misses
+1.3% 6.541e+08 perf-stat.i.dTLB-stores
0.53 ? 8% perf-stat.i.ipc
2936 perf-stat.i.minor-faults
2936 perf-stat.i.page-faults
4604 ? 9% perf-stat.overall.cycles-between-cache-misses
-26.4% 1301105 ? 23% perf-stat.ps.cache-misses
+1.3% 6.524e+08 perf-stat.ps.dTLB-stores
2929 perf-stat.ps.minor-faults
2929 perf-stat.ps.page-faults
0.96 ? 13% perf-profile.calltrace.cycles-pp.exit_to_user_mode_prepare.syscall_exit_to_user_mode.entry_SYSCALL_64_after_hwframe
1.74 ? 14% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.entry_SYSCALL_64_after_hwframe
0.18 ? 8% perf-profile.children.cycles-pp.call_cpuidle
0.18 ? 24% perf-profile.children.cycles-pp.do_user_addr_fault
0.20 ? 18% perf-profile.children.cycles-pp.asm_exc_page_fault
0.18 ? 20% perf-profile.children.cycles-pp.exc_page_fault
0.16 ? 22% perf-profile.children.cycles-pp.__handle_mm_fault
0.17 ? 22% perf-profile.children.cycles-pp.handle_mm_fault
1.24 ? 13% perf-profile.children.cycles-pp.exit_to_user_mode_prepare
0.60 ? 20% perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
2.51 ? 15% perf-profile.children.cycles-pp.syscall_exit_to_user_mode
0.18 ? 8% perf-profile.self.cycles-pp.call_cpuidle
0.10 ? 27% perf-profile.self.cycles-pp.trigger_load_balance
1.12 ? 15% perf-profile.self.cycles-pp.exit_to_user_mode_prepare
0.53 ? 17% perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
2795 ? 5% proc-vmstat.nr_active_anon
187.00 ? 74% proc-vmstat.nr_active_file
62014 proc-vmstat.nr_anon_pages
+3.9% 248685 proc-vmstat.nr_file_pages
64234 proc-vmstat.nr_inactive_anon
8483 proc-vmstat.nr_mapped
5208 ? 3% proc-vmstat.nr_shmem
+3.4% 243282 proc-vmstat.nr_unevictable
2795 ? 5% proc-vmstat.nr_zone_active_anon
187.00 ? 74% proc-vmstat.nr_zone_active_file
64234 proc-vmstat.nr_zone_inactive_anon
+3.4% 243282 proc-vmstat.nr_zone_unevictable
+14.1% 1069639 proc-vmstat.numa_hit
+8.1% 979942 proc-vmstat.numa_local
89696 ? 5% proc-vmstat.numa_other
3827 ? 3% proc-vmstat.pgactivate
+13.8% 1134186 proc-vmstat.pgalloc_normal
+1.3% 1189681 proc-vmstat.pgfault
+12.2% 1157491 proc-vmstat.pgfree
78.25 ? 30% interrupts.CPU11.NMI:Non-maskable_interrupts
78.25 ? 30% interrupts.CPU11.PMI:Performance_monitoring_interrupts
62.75 ? 61% interrupts.CPU12.RES:Rescheduling_interrupts
997.00 ? 82% interrupts.CPU36.CAL:Function_call_interrupts
609.25 ? 19% interrupts.CPU38.CAL:Function_call_interrupts
23198 ?166% interrupts.CPU4.CAL:Function_call_interrupts
719.25 ? 44% interrupts.CPU42.CAL:Function_call_interrupts
545.75 ? 3% interrupts.CPU45.CAL:Function_call_interrupts
79.75 ? 42% interrupts.CPU53.NMI:Non-maskable_interrupts
79.75 ? 42% interrupts.CPU53.PMI:Performance_monitoring_interrupts
94.50 ? 27% interrupts.CPU54.NMI:Non-maskable_interrupts
94.50 ? 27% interrupts.CPU54.PMI:Performance_monitoring_interrupts
65.00 ? 33% interrupts.CPU55.NMI:Non-maskable_interrupts
65.00 ? 33% interrupts.CPU55.PMI:Performance_monitoring_interrupts
90.75 ? 27% interrupts.CPU56.NMI:Non-maskable_interrupts
90.75 ? 27% interrupts.CPU56.PMI:Performance_monitoring_interrupts
78.25 ? 23% interrupts.CPU58.NMI:Non-maskable_interrupts
78.25 ? 23% interrupts.CPU58.PMI:Performance_monitoring_interrupts
1128 ? 85% interrupts.CPU81.CAL:Function_call_interrupts
2179 ? 76% interrupts.CPU90.CAL:Function_call_interrupts
2128 ? 10% interrupts.RES:Rescheduling_interrupts
0.00 sched_debug.cfs_rq:/.exec_clock.avg
0.00 sched_debug.cfs_rq:/.exec_clock.max
0.00 sched_debug.cfs_rq:/.exec_clock.min
0.00 sched_debug.cfs_rq:/.exec_clock.stddev
0.05 ? 9% sched_debug.cpu.nr_running.avg
0.00 sched_debug.cpu.sched_count.avg
0.00 sched_debug.cpu.sched_count.max
0.00 sched_debug.cpu.sched_count.min
0.00 sched_debug.cpu.sched_count.stddev
0.00 sched_debug.cpu.sched_goidle.avg
0.00 sched_debug.cpu.sched_goidle.max
0.00 sched_debug.cpu.sched_goidle.min
0.00 sched_debug.cpu.sched_goidle.stddev
0.00 sched_debug.cpu.ttwu_count.avg
0.00 sched_debug.cpu.ttwu_count.max
0.00 sched_debug.cpu.ttwu_count.min
0.00 sched_debug.cpu.ttwu_count.stddev
0.00 sched_debug.cpu.ttwu_local.avg
0.00 sched_debug.cpu.ttwu_local.max
0.00 sched_debug.cpu.ttwu_local.min
0.00 sched_debug.cpu.ttwu_local.stddev
estimated based on internal Intel analysis and are provided
purposes only. Any difference in system hardware or software
may affect actual performance.

Attachments:

(No filename) (26.41 kB)
config-5.10.0-rc4-00001-ga358d5636cfc (172.75 kB)
job-script (7.93 kB)
job.yaml (5.20 kB)
reproduce (348.00 B)
Download all attachments