Bug#995997: linux-image-5.14.0-2-amd64: Apps issue lots of clock_gettime calls on 5.14, performance drop in some cases.
Seems to be related: 5.14 chooses hpet clocksource, while 5.10 uses tsc:
kernel: Linux version 5.10.0-8-amd64 (debian-kernel@lists.debian.org) (gcc-10 (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2) #1 SMP Debian 5.10.46-4 (2021-08-03)
kernel: tsc: Fast TSC calibration using PIT
kernel: tsc: Detected 2595.027 MHz processor
kernel: clocksource: refined-jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 7645519600211568 ns
kernel: clocksource: hpet: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 133484873504 ns
kernel: clocksource: tsc-early: mask: 0xffffffffffffffff max_cycles: 0x2567e268986, max_idle_ns: 440795272522 ns
kernel: clocksource: jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 7645041785100000 ns
kernel: hpet0: at MMIO 0xfed00000, IRQs 2, 8, 0
kernel: hpet0: 3 comparators, 32-bit 14.318180 MHz counter
kernel: clocksource: Switched to clocksource tsc-early
kernel: clocksource: acpi_pm: mask: 0xffffff max_cycles: 0xffffff, max_idle_ns: 2085701024 ns
kernel: rtc_cmos 00:01: setting system clock to 2021-10-08T21:35:56 UTC (1633728956)
kernel: rtc_cmos 00:01: alarms up to one month, 114 bytes nvram, hpet irqs
kernel: sched_clock: Marking stable (940089949, 397183)->(945335255, -4848123)
kernel: tsc: Refined TSC clocksource calibration: 2615.448 MHz
kernel: clocksource: tsc: mask: 0xffffffffffffffff max_cycles: 0x25b33d90d1e, max_idle_ns: 440795316202 ns
kernel: clocksource: Switched to clocksource tsc
kernel: Linux version 5.14.0-2-amd64 (debian-kernel@lists.debian.org) (gcc-10 (Debian 10.3.0-11) 10.3.0, GNU ld (GNU Binutils for Debian) 2.37) #1 SMP Debian 5.14.9-2 (2021-10-03)
kernel: tsc: Fast TSC calibration using PIT
kernel: tsc: Detected 2595.027 MHz processor
kernel: clocksource: refined-jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 7645519600211568 ns
kernel: clocksource: hpet: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 133484873504 ns
kernel: clocksource: tsc-early: mask: 0xffffffffffffffff max_cycles: 0x2567e268986, max_idle_ns: 440795272522 ns
kernel: clocksource: jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 7645041785100000 ns
kernel: hpet0: at MMIO 0xfed00000, IRQs 2, 8, 0
kernel: hpet0: 3 comparators, 32-bit 14.318180 MHz counter
kernel: clocksource: Switched to clocksource tsc-early
kernel: clocksource: acpi_pm: mask: 0xffffff max_cycles: 0xffffff, max_idle_ns: 2085701024 ns
kernel: rtc_cmos 00:01: setting system clock to 2021-10-08T21:38:57 UTC (1633729137)
kernel: rtc_cmos 00:01: alarms up to one month, 114 bytes nvram, hpet irqs
kernel: sched_clock: Marking stable (1012272366, 406541)->(1016715343, -4036436)
kernel: tsc: Refined TSC clocksource calibration: 2615.168 MHz
kernel: clocksource: tsc: mask: 0xffffffffffffffff max_cycles: 0x25b234d5bf1, max_idle_ns: 440795256376 ns
kernel: clocksource: Switched to clocksource tsc
kernel: clocksource: timekeeping watchdog on CPU3: Marking clocksource 'tsc' as unstable because the skew is too large:
kernel: clocksource: 'hpet' wd_nsec: 515995748 wd_now: 1bec100 wd_last: 14e0528 mask: ffffffff
kernel: clocksource: 'tsc' cs_nsec: 512040924 cs_now: 547cae902 cs_last: 4f7fa46c4 mask: ffffffffffffffff
kernel: clocksource: 'tsc' is current clocksource.
kernel: tsc: Marking TSC unstable due to clocksource watchdog
kernel: TSC found unstable after boot, most likely due to broken BIOS. Use 'tsc=unstable'.
kernel: sched_clock: Marking unstable (2098965976, 406565)<-(2103408384, -4036436)
kernel: clocksource: Checking clocksource tsc synchronization from CPU 1 to CPUs 0,2,4,6.
kernel: clocksource: Switched to clocksource hpet
Reply to: