clocksource 解释

在 @2018-01-14 版更改: 创建

内核日志：

clocksource: timekeeping watchdog on CPU7: Marking clocksource 'tsc' as unstable because the skew is too large:
clocksource:                       'hpet' wd_now: ec7fd07a wd_last: f6240e67 mask: ffffffff
clocksource:                       'tsc' cs_now: db522085db96 cs_last: 9d30c4ea512c mask: ffffffffffffffff
clocksource: Switched to clocksource hpet

If system clcok source is hpet, then gettimeofday will do syscall gettimeofday, not for clocksource tsc. So we should add gettimeofday to seccomp too, otherwise:

[WARN/seccomp] invalid syscall attempted: gettimeofday(96)
[CRIT/seccomp] invalid syscall not allowed: stop here

如下测试代码演示差异：

#include <stdio.h>
#include <sys/time.h>

int main(int argc, char **argv)
{
        struct timeval tv;

        gettimeofday(&tv, NULL);
        printf("%ld\n", tv.tv_sec);
        return 0;
}

下面是不同 clocksource 时 strace 的差异：

$ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
hpet
$ strace -f ./a.out
execve("./a.out", ["./a.out"], [/* 39 vars */]) = 0
brk(NULL)                               = 0x559844c55000
…
gettimeofday({tv_sec=1512111369, tv_usec=671919}, NULL) = 0
fstat(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 0), ...}) = 0
brk(NULL)                               = 0x559844c55000
brk(0x559844c76000)                     = 0x559844c76000
write(1, "1512111369\n", 111512111369
)            = 11
exit_group(0)                           = ?
+++ exited with 0 +++

$ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
tsc
$ strace -f ./a.out
execve("./a.out", ["./a.out"], [/* 47 vars */]) = 0
brk(NULL)                               = 0x7f5b0b28e000
…
fstat(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 96), ...}) = 0
brk(NULL)                               = 0x7f5b0b28e000
brk(0x7f5b0b2af000)                     = 0x7f5b0b2af000
write(1, "1512111377\n", 111512111377
)            = 11
exit_group(0)                           = ?
+++ exited with 0 +++