Re: [LKP] Re: [x86/signal] 3aac3ebea0: will-it-scale.per_thread_ops -11.9% regression

From: Carel Si
Date: Wed Dec 08 2021 - 21:33:13 EST

Hi Thomas,

On Tue, Dec 07, 2021 at 02:38:34PM +0100, Thomas Gleixner wrote:
> Hi!
>
> On Tue, Dec 07 2021 at 09:21, kernel test robot wrote:
>
> > (please be noted we made some further analysis before reporting out,
> > and thought it's likely the regression is related with the extra spinlock
> > introducded by enalbling DYNAMIC_SIGFRAME. below is the full report.)
> >
> > FYI, we noticed a -11.9% regression of will-it-scale.per_thread_ops due to commit:
>
> Does that use sigaltstack() ?
>
> > 1bdda24c4af64cd2 3aac3ebea08f2d342364f827c89
> > ---------------- ---------------------------
> > %stddev %change %stddev
> > \ | \
> > 754824 ± 2% -11.9% 664668 ± 2% will-it-scale.16.threads
> > 47176 ± 2% -11.9% 41541 ± 2% will-it-scale.per_thread_ops
> > 754824 ± 2% -11.9% 664668 ± 2% will-it-scale.workload
> > 1422782 ± 8% +3.3e+05 1751520 ± 12% syscalls.sys_getpid.noise.5%
>
> Somehow the printout got confused ...
>
> > 1.583e+10 -2.1% 1.55e+10 perf-stat.i.instructions
> > 6328594 ± 2% +11.1% 7032338 ± 2% perf-stat.overall.path-length
> > 1.578e+10 -2.1% 1.545e+10 perf-stat.ps.instructions
> > 4.774e+12 -2.2% 4.671e+12 perf-stat.total.instructions
> > 0.00 +6.3 6.33 ± 8% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.do_sigaltstack.restore_altstack.__x64_sys_rt_sigreturn
> > 0.00 +6.5 6.51 ± 8% perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.do_sigaltstack.restore_altstack.__x64_sys_rt_sigreturn.do_syscall_64
> > 0.00 +6.6 6.58 ± 8% perf-profile.calltrace.cycles-pp.do_sigaltstack.restore_altstack.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe
> > 0.00 +6.6 6.62 ± 8% perf-profile.calltrace.cycles-pp.restore_altstack.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
> > 0.00 +6.9 6.88 ± 9% perf-profile.calltrace.cycles-pp.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
> > 7.99 ± 12% +6.0 14.00 ± 9% perf-profile.children.cycles-pp.__x64_sys_rt_sigreturn
> > 0.05 ± 44% +6.6 6.62 ± 8% perf-profile.children.cycles-pp.restore_altstack
> > 0.00 +6.6 6.58 ± 8% perf-profile.children.cycles-pp.do_sigaltstack
>
> It looks like it does. The problem is that sighand->lock is process
> wide.
>
> Can you test whether the below cures it?
>

We applied your patch upon mainline commit 2a987e6502 ("Merge tag
'perf-tools-fixes-for-v5.16-2021-12-07' of
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux"), it will bring 9%
improvement. Thanks.

=========================================================================================
compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/testcase/ucode:
gcc-9/performance/x86_64-rhel-8.3/thread/16/debian-10.4-x86_64-20200603.cgz/lkp-hsw-4ex1/signal1/will-it-scale/0x16

commit:
2a987e6502 ("Merge tag 'perf-tools-fixes-for-v5.16-2021-12-07' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux")
fceec50b60 ("fixup-for-2a987e6502")

2a987e65025e2b79 fceec50b600c90a3a3ac3406c03
---------------- ---------------------------
%stddev %change %stddev
\ | \
660062 ± 2% +9.0% 719344 will-it-scale.16.threads
41253 ± 2% +9.0% 44958 will-it-scale.per_thread_ops
660062 ± 2% +9.0% 719344 will-it-scale.workload
38126 ± 35% -6.2% 35753 ± 12% syscalls.sys_getpid.max
347.25 -0.4% 346.00 syscalls.sys_getpid.med
324.00 -0.2% 323.50 syscalls.sys_getpid.min
866263 ± 7% -44781.2 821482 ± 10% syscalls.sys_getpid.noise.100%
1916161 ± 5% -64520.2 1851640 ± 4% syscalls.sys_getpid.noise.2%
1268029 ± 5% -50154.5 1217875 ± 7% syscalls.sys_getpid.noise.25%
1722829 ± 7% -1e+05 1621521 ± 5% syscalls.sys_getpid.noise.5%
1167288 ± 5% -40972.4 1126316 ± 8% syscalls.sys_getpid.noise.50%
1072219 ± 6% -53120.8 1019098 ± 8% syscalls.sys_getpid.noise.75%
54168 ± 39% -38.5% 33334 ± 15% syscalls.sys_gettid.max
333.75 -0.2% 333.00 syscalls.sys_gettid.med
315.75 -0.2% 315.00 syscalls.sys_gettid.min
923814 ± 13% -1.3e+05 795012 ± 11% syscalls.sys_gettid.noise.100%
1909235 ± 6% -1.2e+05 1788745 ± 5% syscalls.sys_gettid.noise.2%
1254536 ± 10% -1.2e+05 1134475 ± 7% syscalls.sys_gettid.noise.25%
1664843 ± 8% -1.2e+05 1544153 ± 6% syscalls.sys_gettid.noise.5%
1209931 ± 10% -1.2e+05 1092160 ± 7% syscalls.sys_gettid.noise.50%
1120212 ± 10% -1.2e+05 1004727 ± 8% syscalls.sys_gettid.noise.75%
3.64e+09 ± 8% +83.1% 6.666e+09 ± 92% syscalls.sys_read.max
1837 ± 2% +3.6% 1902 syscalls.sys_read.med
669.75 ± 3% +1.7% 681.33 ± 4% syscalls.sys_read.min
8.308e+11 +2.5e+10 8.556e+11 ± 8% syscalls.sys_read.noise.100%
8.308e+11 +2.5e+10 8.557e+11 ± 8% syscalls.sys_read.noise.2%
8.308e+11 +2.5e+10 8.557e+11 ± 8% syscalls.sys_read.noise.25%
8.308e+11 +2.5e+10 8.557e+11 ± 8% syscalls.sys_read.noise.5%
8.308e+11 +2.5e+10 8.557e+11 ± 8% syscalls.sys_read.noise.50%
8.308e+11 +2.5e+10 8.557e+11 ± 8% syscalls.sys_read.noise.75%
27686929 ±172% -47.1% 14660048 ±219% syscalls.sys_rt_sigprocmask.max
7603 ± 2% +1.1% 7689 syscalls.sys_rt_sigprocmask.med
554.75 ± 5% +4.1% 577.67 ± 9% syscalls.sys_rt_sigprocmask.min
59550208 ±117% -2.3e+07 36678292 ±117% syscalls.sys_rt_sigprocmask.noise.100%
99385689 ± 69% -2.4e+07 75649798 ± 55% syscalls.sys_rt_sigprocmask.noise.2%
70154045 ± 98% -2.3e+07 47366143 ± 89% syscalls.sys_rt_sigprocmask.noise.25%
96889781 ± 71% -2.4e+07 72970986 ± 57% syscalls.sys_rt_sigprocmask.noise.5%
65994619 ±105% -2.2e+07 43517676 ± 97% syscalls.sys_rt_sigprocmask.noise.50%
62923250 ±110% -2.3e+07 40273591 ±106% syscalls.sys_rt_sigprocmask.noise.75%
28208603 ±171% -32.8% 18960706 ±220% syscalls.sys_tgkill.max
7711 ± 2% +4.8% 8080 syscalls.sys_tgkill.med
1337 ± 2% +1.9% 1362 ± 2% syscalls.sys_tgkill.min
51870512 ±109% -9.8e+06 42069157 ±126% syscalls.sys_tgkill.noise.100%
1.015e+08 ± 54% -1.5e+07 86618122 ± 61% syscalls.sys_tgkill.noise.2%
68743216 ± 80% -1.3e+07 55954685 ± 95% syscalls.sys_tgkill.noise.25%
99442048 ± 55% -1.5e+07 83995198 ± 63% syscalls.sys_tgkill.noise.5%
59229382 ± 95% -9.1e+06 50108976 ±106% syscalls.sys_tgkill.noise.50%
55423780 ±101% -9.6e+06 45845571 ±116% syscalls.sys_tgkill.noise.75%
7.13 ± 23% +7.6% 7.68 ± 7% perf-stat.i.MPKI
3.518e+09 -0.1% 3.517e+09 perf-stat.i.branch-instructions
1.02 ± 18% +0.1 1.11 ± 6% perf-stat.i.branch-miss-rate%
35924567 ± 18% +9.0% 39155875 ± 6% perf-stat.i.branch-misses
0.25 ± 42% -0.0 0.21 ± 26% perf-stat.i.cache-miss-rate%
254920 ± 8% -0.5% 253744 ± 15% perf-stat.i.cache-misses
1.105e+08 ± 24% +8.2% 1.196e+08 ± 7% perf-stat.i.cache-references
1924 -0.6% 1912 perf-stat.i.context-switches
3.39 ± 2% +0.0% 3.39 perf-stat.i.cpi
144011 +0.0% 144013 perf-stat.i.cpu-clock
5.251e+10 ± 3% +0.6% 5.283e+10 perf-stat.i.cpu-cycles
149.17 -0.3% 148.80 perf-stat.i.cpu-migrations
238223 ± 10% +3.4% 246267 ± 11% perf-stat.i.cycles-between-cache-misses
0.17 ± 17% +0.0 0.19 ± 6% perf-stat.i.dTLB-load-miss-rate%
7107283 ± 17% +12.4% 7986549 ± 5% perf-stat.i.dTLB-load-misses
4.219e+09 +1.7% 4.29e+09 perf-stat.i.dTLB-loads
0.28 +0.0 0.28 perf-stat.i.dTLB-store-miss-rate%
4776002 ± 7% +9.6% 5232957 ± 2% perf-stat.i.dTLB-store-misses
1.707e+09 ± 8% +8.5% 1.852e+09 ± 3% perf-stat.i.dTLB-stores
76.72 ± 5% +2.6 79.30 ± 2% perf-stat.i.iTLB-load-miss-rate%
7312020 ± 9% +6.8% 7811089 ± 2% perf-stat.i.iTLB-load-misses
2197456 ± 17% -7.8% 2024986 ± 6% perf-stat.i.iTLB-loads
1.551e+10 +0.6% 1.56e+10 perf-stat.i.instructions
2153 ± 8% -6.4% 2016 ± 3% perf-stat.i.instructions-per-iTLB-miss
0.30 ± 2% -0.0% 0.30 perf-stat.i.ipc
1.01 ± 3% +1.4% 1.02 ± 3% perf-stat.i.major-faults
0.36 ± 3% +0.6% 0.37 perf-stat.i.metric.GHz
811.17 ± 22% +7.8% 874.34 ± 6% perf-stat.i.metric.K/sec
65.59 +2.3% 67.08 perf-stat.i.metric.M/sec
3816 -0.1% 3812 perf-stat.i.minor-faults
94.33 +0.6 94.95 perf-stat.i.node-load-miss-rate%
155138 ± 18% -2.8% 150833 ± 20% perf-stat.i.node-load-misses
17825 ± 11% -18.6% 14509 ± 6% perf-stat.i.node-loads
61.65 ± 7% +3.6 65.20 ± 4% perf-stat.i.node-store-miss-rate%
53253 ± 11% -1.7% 52341 ± 2% perf-stat.i.node-store-misses
36553 ± 18% -12.1% 32126 ± 19% perf-stat.i.node-stores
3817 -0.1% 3813 perf-stat.i.page-faults
144011 +0.0% 144013 perf-stat.i.task-clock
7.12 ± 23% +7.7% 7.66 ± 7% perf-stat.overall.MPKI
1.02 ± 17% +0.1 1.11 ± 6% perf-stat.overall.branch-miss-rate%
0.26 ± 43% -0.0 0.22 ± 26% perf-stat.overall.cache-miss-rate%
3.39 ± 2% -0.0% 3.39 perf-stat.overall.cpi
207309 ± 10% +2.3% 211978 ± 12% perf-stat.overall.cycles-between-cache-misses
0.17 ± 17% +0.0 0.19 ± 6% perf-stat.overall.dTLB-load-miss-rate%
0.28 +0.0 0.28 perf-stat.overall.dTLB-store-miss-rate%
76.78 ± 5% +2.6 79.39 ± 2% perf-stat.overall.iTLB-load-miss-rate%
2139 ± 8% -6.5% 2000 ± 3% perf-stat.overall.instructions-per-iTLB-miss
0.30 ± 2% -0.1% 0.30 perf-stat.overall.ipc
89.49 +1.5 90.98 perf-stat.overall.node-load-miss-rate%
59.42 ± 7% +2.8 62.26 ± 6% perf-stat.overall.node-store-miss-rate%
7090142 -7.8% 6537840 perf-stat.overall.path-length
3.507e+09 -0.1% 3.505e+09 perf-stat.ps.branch-instructions
35849751 ± 18% +9.0% 39063558 ± 6% perf-stat.ps.branch-misses
254856 ± 8% -0.5% 253483 ± 15% perf-stat.ps.cache-misses
1.101e+08 ± 24% +8.2% 1.192e+08 ± 7% perf-stat.ps.cache-references
1916 -0.6% 1904 perf-stat.ps.context-switches
143524 -0.0% 143524 perf-stat.ps.cpu-clock
5.234e+10 ± 3% +0.6% 5.265e+10 perf-stat.ps.cpu-cycles
148.93 -0.3% 148.51 perf-stat.ps.cpu-migrations
7083441 ± 17% +12.4% 7959311 ± 5% perf-stat.ps.dTLB-load-misses
4.206e+09 +1.7% 4.276e+09 perf-stat.ps.dTLB-loads
4760073 ± 7% +9.6% 5215393 ± 2% perf-stat.ps.dTLB-store-misses
1.702e+09 ± 8% +8.5% 1.846e+09 ± 3% perf-stat.ps.dTLB-stores
7286389 ± 9% +6.8% 7783435 ± 2% perf-stat.ps.iTLB-load-misses
2190584 ± 17% -7.9% 2018547 ± 6% perf-stat.ps.iTLB-loads
1.546e+10 +0.6% 1.555e+10 perf-stat.ps.instructions
1.01 ± 3% +1.6% 1.02 ± 3% perf-stat.ps.major-faults
3788 -0.1% 3784 perf-stat.ps.minor-faults
154954 ± 18% -2.8% 150666 ± 20% perf-stat.ps.node-load-misses
17958 ± 10% -18.8% 14577 ± 6% perf-stat.ps.node-loads
53120 ± 11% -1.7% 52211 ± 2% perf-stat.ps.node-store-misses
36482 ± 18% -12.1% 32073 ± 19% perf-stat.ps.node-stores
3789 -0.1% 3785 perf-stat.ps.page-faults
143524 -0.0% 143524 perf-stat.ps.task-clock
4.678e+12 +0.5% 4.703e+12 perf-stat.total.instructions
6.28 ± 10% -6.3 0.00 perf-profile.calltrace.cycles-pp.restore_altstack.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
6.25 ± 10% -6.2 0.00 perf-profile.calltrace.cycles-pp.do_sigaltstack.restore_altstack.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe
6.18 ± 10% -6.2 0.00 perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.do_sigaltstack.restore_altstack.__x64_sys_rt_sigreturn.do_syscall_64
6.01 ± 10% -6.0 0.00 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.do_sigaltstack.restore_altstack.__x64_sys_rt_sigreturn
6.55 ± 10% -5.9 0.65 ± 10% perf-profile.calltrace.cycles-pp.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
38.33 ± 15% -3.5 34.84 ± 17% perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.cpuidle_enter.do_idle.cpu_startup_entry
40.09 ± 13% -3.4 36.70 ± 14% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.cpuidle_enter.do_idle.cpu_startup_entry.start_secondary
40.26 ± 13% -3.4 36.88 ± 14% perf-profile.calltrace.cycles-pp.cpuidle_enter.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64_no_verify
41.08 ± 13% -3.3 37.76 ± 13% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64_no_verify
41.09 ± 13% -3.3 37.78 ± 13% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.secondary_startup_64_no_verify
41.09 ± 13% -3.3 37.78 ± 13% perf-profile.calltrace.cycles-pp.start_secondary.secondary_startup_64_no_verify
41.76 ± 13% -3.3 38.44 ± 14% perf-profile.calltrace.cycles-pp.secondary_startup_64_no_verify
0.29 ±101% -0.1 0.19 ±142% perf-profile.calltrace.cycles-pp.tick_sched_handle.tick_sched_timer.__hrtimer_run_queues.hrtimer_interrupt.__sysvec_apic_timer_interrupt
0.14 ±173% -0.1 0.09 ±223% perf-profile.calltrace.cycles-pp.update_process_times.tick_sched_handle.tick_sched_timer.__hrtimer_run_queues.hrtimer_interrupt
0.45 ± 60% +0.0 0.48 ± 46% perf-profile.calltrace.cycles-pp.tick_sched_timer.__hrtimer_run_queues.hrtimer_interrupt.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt
0.60 ±106% +0.0 0.63 ±103% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_kernel.secondary_startup_64_no_verify
0.60 ±106% +0.0 0.63 ±103% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64_no_verify
0.60 ±106% +0.0 0.63 ±103% perf-profile.calltrace.cycles-pp.cpuidle_enter.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64_no_verify
0.60 ±106% +0.0 0.63 ±103% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.cpuidle_enter.do_idle.cpu_startup_entry.start_kernel
0.60 ±106% +0.0 0.63 ±103% perf-profile.calltrace.cycles-pp.start_kernel.secondary_startup_64_no_verify
1.34 ± 18% +0.0 1.39 ± 13% perf-profile.calltrace.cycles-pp.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state.cpuidle_enter
1.28 ± 18% +0.0 1.33 ± 13% perf-profile.calltrace.cycles-pp.hrtimer_interrupt.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state
0.88 ± 23% +0.1 0.94 ± 12% perf-profile.calltrace.cycles-pp.__hrtimer_run_queues.hrtimer_interrupt.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt
1.96 ± 17% +0.1 2.03 ± 13% perf-profile.calltrace.cycles-pp.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state.cpuidle_enter.do_idle
0.46 ± 59% +0.1 0.54 ± 46% perf-profile.calltrace.cycles-pp.menu_select.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64_no_verify
2.15 ± 16% +0.1 2.23 ± 13% perf-profile.calltrace.cycles-pp.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state.cpuidle_enter.do_idle.cpu_startup_entry
0.00 +0.1 0.08 ±223% perf-profile.calltrace.cycles-pp.__set_task_blocked.__set_current_blocked.sigprocmask.__x64_sys_rt_sigprocmask.do_syscall_64
0.60 ± 8% +0.1 0.72 ± 7% perf-profile.calltrace.cycles-pp.ring_buffer_lock_reserve.trace_buffer_lock_reserve.ftrace_syscall_exit.syscall_exit_to_user_mode_prepare.syscall_exit_to_user_mode
0.90 ± 8% +0.1 1.04 ± 9% perf-profile.calltrace.cycles-pp.__entry_text_start.raise
0.66 ± 8% +0.1 0.79 ± 7% perf-profile.calltrace.cycles-pp.trace_buffer_lock_reserve.ftrace_syscall_exit.syscall_exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64
0.25 ±100% +0.1 0.40 ± 70% perf-profile.calltrace.cycles-pp.__send_signal.do_send_sig_info.do_send_specific.do_tkill.__x64_sys_tgkill
0.91 ± 8% +0.2 1.07 ± 10% perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.raise
40.06 ± 9% +0.2 40.22 ± 9% perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
1.02 ± 7% +0.2 1.20 ± 8% perf-profile.calltrace.cycles-pp.ftrace_syscall_enter.syscall_trace_enter.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
1.00 ± 9% +0.2 1.18 ± 8% perf-profile.calltrace.cycles-pp.ftrace_syscall_exit.syscall_exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe
40.17 ± 9% +0.2 40.35 ± 9% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.raise
1.11 ± 8% +0.2 1.30 ± 9% perf-profile.calltrace.cycles-pp.syscall_trace_enter.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
1.08 ± 9% +0.2 1.27 ± 7% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
0.27 ±100% +0.3 0.59 ± 9% perf-profile.calltrace.cycles-pp.trace_buffer_lock_reserve.ftrace_syscall_enter.syscall_trace_enter.do_syscall_64.entry_SYSCALL_64_after_hwframe
0.13 ±173% +0.3 0.47 ± 45% perf-profile.calltrace.cycles-pp.__rb_reserve_next.ring_buffer_lock_reserve.trace_buffer_lock_reserve.ftrace_syscall_exit.syscall_exit_to_user_mode_prepare
0.00 +0.4 0.36 ± 70% perf-profile.calltrace.cycles-pp.ring_buffer_lock_reserve.trace_buffer_lock_reserve.ftrace_syscall_enter.syscall_trace_enter.do_syscall_64
42.28 ± 9% +0.5 42.82 ± 8% perf-profile.calltrace.cycles-pp.raise
6.59 ± 9% +1.0 7.56 ± 8% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.get_signal.arch_do_signal_or_restart.exit_to_user_mode_prepare
6.76 ± 9% +1.0 7.78 ± 8% perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.get_signal.arch_do_signal_or_restart.exit_to_user_mode_prepare.syscall_exit_to_user_mode
7.17 ± 9% +1.1 8.26 ± 8% perf-profile.calltrace.cycles-pp.get_signal.arch_do_signal_or_restart.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64
6.45 ± 9% +1.1 7.57 ± 9% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.__set_current_blocked.signal_setup_done.arch_do_signal_or_restart
6.12 ± 10% +1.1 7.26 ± 8% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.__set_current_blocked.__x64_sys_rt_sigreturn.do_syscall_64
6.61 ± 9% +1.2 7.76 ± 9% perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.__set_current_blocked.signal_setup_done.arch_do_signal_or_restart.exit_to_user_mode_prepare
7.63 ± 9% +1.2 8.79 ± 8% perf-profile.calltrace.cycles-pp.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
6.30 ± 10% +1.2 7.50 ± 8% perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.__set_current_blocked.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe
6.88 ± 9% +1.2 8.08 ± 9% perf-profile.calltrace.cycles-pp.__set_current_blocked.signal_setup_done.arch_do_signal_or_restart.exit_to_user_mode_prepare.syscall_exit_to_user_mode
6.89 ± 9% +1.2 8.10 ± 9% perf-profile.calltrace.cycles-pp.signal_setup_done.arch_do_signal_or_restart.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64
6.56 ± 9% +1.2 7.79 ± 8% perf-profile.calltrace.cycles-pp.__set_current_blocked.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe.__restore_rt
7.01 ± 9% +1.2 8.24 ± 9% perf-profile.calltrace.cycles-pp.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.handler
6.71 ± 9% +1.2 7.95 ± 8% perf-profile.calltrace.cycles-pp.__x64_sys_rt_sigreturn.do_syscall_64.entry_SYSCALL_64_after_hwframe.__restore_rt
6.96 ± 9% +1.3 8.26 ± 8% perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.__restore_rt
6.97 ± 10% +1.3 8.27 ± 8% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.__restore_rt
7.02 ± 10% +1.3 8.32 ± 8% perf-profile.calltrace.cycles-pp.__restore_rt
6.04 ± 11% +1.3 7.36 ± 9% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__lock_task_sighand.do_send_sig_info.do_send_specific
7.60 ± 9% +1.3 8.95 ± 9% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.handler
7.59 ± 9% +1.4 8.94 ± 9% perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.handler
7.58 ± 9% +1.4 8.93 ± 9% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.handler
6.23 ± 10% +1.4 7.59 ± 9% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__lock_task_sighand.do_send_sig_info.do_send_specific.do_tkill
6.23 ± 11% +1.4 7.60 ± 9% perf-profile.calltrace.cycles-pp.__lock_task_sighand.do_send_sig_info.do_send_specific.do_tkill.__x64_sys_tgkill
8.27 ± 9% +1.4 9.72 ± 9% perf-profile.calltrace.cycles-pp.handler
6.81 ± 10% +1.5 8.29 ± 9% perf-profile.calltrace.cycles-pp.do_send_sig_info.do_send_specific.do_tkill.__x64_sys_tgkill.do_syscall_64
7.12 ± 10% +1.6 8.68 ± 9% perf-profile.calltrace.cycles-pp.do_send_specific.do_tkill.__x64_sys_tgkill.do_syscall_64.entry_SYSCALL_64_after_hwframe
7.18 ± 10% +1.6 8.75 ± 9% perf-profile.calltrace.cycles-pp.do_tkill.__x64_sys_tgkill.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
7.20 ± 10% +1.6 8.77 ± 9% perf-profile.calltrace.cycles-pp.__x64_sys_tgkill.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
11.57 ± 9% +1.6 13.20 ± 8% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
14.57 ± 9% +2.4 16.95 ± 9% perf-profile.calltrace.cycles-pp.arch_do_signal_or_restart.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe
12.32 ± 9% +2.4 14.75 ± 9% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.__set_current_blocked.sigprocmask.__x64_sys_rt_sigprocmask
12.70 ± 9% +2.5 15.19 ± 9% perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.__set_current_blocked.sigprocmask.__x64_sys_rt_sigprocmask.do_syscall_64
13.20 ± 9% +2.6 15.79 ± 9% perf-profile.calltrace.cycles-pp.__set_current_blocked.sigprocmask.__x64_sys_rt_sigprocmask.do_syscall_64.entry_SYSCALL_64_after_hwframe
13.22 ± 9% +2.6 15.82 ± 9% perf-profile.calltrace.cycles-pp.sigprocmask.__x64_sys_rt_sigprocmask.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
13.40 ± 9% +2.6 16.02 ± 9% perf-profile.calltrace.cycles-pp.__x64_sys_rt_sigprocmask.do_syscall_64.entry_SYSCALL_64_after_hwframe.raise
6.28 ± 10% -5.9 0.34 ± 10% perf-profile.children.cycles-pp.restore_altstack
6.25 ± 10% -5.9 0.31 ± 11% perf-profile.children.cycles-pp.do_sigaltstack



> Not pretty, but that's what I came up with for now.
>
> Thanks,
>
> tglx
> ---
> --- a/arch/x86/kernel/fpu/core.c
> +++ b/arch/x86/kernel/fpu/core.c
> @@ -457,10 +457,10 @@ static inline void fpu_inherit_perms(str
> if (fpu_state_size_dynamic()) {
> struct fpu *src_fpu = &current->group_leader->thread.fpu;
>
> - spin_lock_irq(&current->sighand->siglock);
> + read_lock(&current->sighand->sigaltstack_lock);
> /* Fork also inherits the permissions of the parent */
> dst_fpu->perm = src_fpu->perm;
> - spin_unlock_irq(&current->sighand->siglock);
> + read_unlock(&current->sighand->sigaltstack_lock);
> }
> }
>
> --- a/arch/x86/kernel/fpu/xstate.c
> +++ b/arch/x86/kernel/fpu/xstate.c
> @@ -1582,17 +1582,22 @@ static int validate_sigaltstack(unsigned
> {
> struct task_struct *thread, *leader = current->group_leader;
> unsigned long framesize = get_sigframe_size();
> + int ret = 0;
>
> - lockdep_assert_held(&current->sighand->siglock);
> + lockdep_assert_held_write(&current->sighand->sigaltstack_lock);
>
> /* get_sigframe_size() is based on fpu_user_cfg.max_size */
> framesize -= fpu_user_cfg.max_size;
> framesize += usize;
> + read_lock(&tasklist_lock);
> for_each_thread(leader, thread) {
> - if (thread->sas_ss_size && thread->sas_ss_size < framesize)
> - return -ENOSPC;
> + if (thread->sas_ss_size && thread->sas_ss_size < framesize) {
> + ret = -ENOSPC;
> + break;
> + }
> }
> - return 0;
> + read_unlock(&tasklist_lock);
> + return ret;
> }
>
> static int __xstate_request_perm(u64 permitted, u64 requested)
> @@ -1627,7 +1632,7 @@ static int __xstate_request_perm(u64 per
>
> /* Pairs with the READ_ONCE() in xstate_get_group_perm() */
> WRITE_ONCE(fpu->perm.__state_perm, requested);
> - /* Protected by sighand lock */
> + /* Protected by sighand::sigaltstack_lock */
> fpu->perm.__state_size = ksize;
> fpu->perm.__user_state_size = usize;
> return ret;
> @@ -1666,10 +1671,10 @@ static int xstate_request_perm(unsigned
> return 0;
>
> /* Protect against concurrent modifications */
> - spin_lock_irq(&current->sighand->siglock);
> + write_lock(&current->sighand->sigaltstack_lock);
> permitted = xstate_get_host_group_perm();
> ret = __xstate_request_perm(permitted, requested);
> - spin_unlock_irq(&current->sighand->siglock);
> + write_unlock(&current->sighand->sigaltstack_lock);
> return ret;
> }
>
> @@ -1685,11 +1690,11 @@ int xfd_enable_feature(u64 xfd_err)
> }
>
> /* Protect against concurrent modifications */
> - spin_lock_irq(&current->sighand->siglock);
> + read_lock(&current->sighand->sigaltstack_lock);
>
> /* If not permitted let it die */
> if ((xstate_get_host_group_perm() & xfd_event) != xfd_event) {
> - spin_unlock_irq(&current->sighand->siglock);
> + read_unlock(&current->sighand->sigaltstack_lock);
> return -EPERM;
> }
>
> @@ -1702,7 +1707,7 @@ int xfd_enable_feature(u64 xfd_err)
> * another task, the retrieved buffer sizes are valid for the
> * currently requested feature(s).
> */
> - spin_unlock_irq(&current->sighand->siglock);
> + read_unlock(&current->sighand->sigaltstack_lock);
>
> /*
> * Try to allocate a new fpstate. If that fails there is no way
> --- a/arch/x86/kernel/signal.c
> +++ b/arch/x86/kernel/signal.c
> @@ -939,17 +939,19 @@ static int __init strict_sas_size(char *
> * the task has permissions to use dynamic features. Tasks which have no
> * permission are checked against the size of the non-dynamic feature set
> * if strict checking is enabled. This avoids forcing all tasks on the
> - * system to allocate large sigaltstacks even if they are never going
> - * to use a dynamic feature. As this is serialized via sighand::siglock
> - * any permission request for a dynamic feature either happened already
> - * or will see the newly install sigaltstack size in the permission checks.
> + * system to allocate large sigaltstacks even if they are never going to
> + * use a dynamic feature.
> + *
> + * As this is serialized via sighand::sigaltstack_lock any permission
> + * request for a dynamic feature either happened already or will see the
> + * newly install sigaltstack size in the permission checks.
> */
> bool sigaltstack_size_valid(size_t ss_size)
> {
> unsigned long fsize = max_frame_size - fpu_default_state_size;
> u64 mask;
>
> - lockdep_assert_held(&current->sighand->siglock);
> + lockdep_assert_held_read(&current->sighand->sigaltstack_lock);
>
> if (!fpu_state_size_dynamic() && !strict_sigaltstack_size)
> return true;
> --- a/include/linux/sched/signal.h
> +++ b/include/linux/sched/signal.h
> @@ -19,6 +19,9 @@
>
> struct sighand_struct {
> spinlock_t siglock;
> +#ifdef CONFIG_DYNAMIC_SIGFRAME
> + rwlock_t sigaltstack_lock;
> +#endif
> refcount_t count;
> wait_queue_head_t signalfd_wqh;
> struct k_sigaction action[_NSIG];
> --- a/init/init_task.c
> +++ b/init/init_task.c
> @@ -48,6 +48,9 @@ static struct sighand_struct init_sighan
> .action = { { { .sa_handler = SIG_DFL, } }, },
> .siglock = __SPIN_LOCK_UNLOCKED(init_sighand.siglock),
> .signalfd_wqh = __WAIT_QUEUE_HEAD_INITIALIZER(init_sighand.signalfd_wqh),
> +#ifdef CONFIG_DYNAMIC_SIGFRAME
> + .sigaltstack_lock = __RW_LOCK_UNLOCKED(init_sighand.sigaltstack_lock),
> +#endif
> };
>
> #ifdef CONFIG_SHADOW_CALL_STACK
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -2900,6 +2900,9 @@ static void sighand_ctor(void *data)
>
> spin_lock_init(&sighand->siglock);
> init_waitqueue_head(&sighand->signalfd_wqh);
> +#ifdef CONFIG_DYNAMIC_SIGFRAME
> + rwlock_init(&sighand->sigaltstack_lock);
> +#endif
> }
>
> void __init proc_caches_init(void)
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -4141,15 +4141,15 @@ int do_sigaction(int sig, struct k_sigac
>
> #ifdef CONFIG_DYNAMIC_SIGFRAME
> static inline void sigaltstack_lock(void)
> - __acquires(&current->sighand->siglock)
> + __acquires(&current->sighand->sigaltstack_lock)
> {
> - spin_lock_irq(&current->sighand->siglock);
> + read_lock(&current->sighand->sigaltstack_lock);
> }
>
> static inline void sigaltstack_unlock(void)
> - __releases(&current->sighand->siglock)
> + __releases(&current->sighand->sigaltstack_lock)
> {
> - spin_unlock_irq(&current->sighand->siglock);
> + read_unlock(&current->sighand->sigaltstack_lock);
> }
> #else
> static inline void sigaltstack_lock(void) { }
> _______________________________________________
> LKP mailing list -- lkp@xxxxxxxxxxxx
> To unsubscribe send an email to lkp-leave@xxxxxxxxxxxx