[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAHmME9pZpJj3zL3nEmz6CSouZs8pmt4D1VO26Zv8imVEPuJFeA@mail.gmail.com>
Date: Tue, 18 Jun 2019 19:34:23 +0200
From: "Jason A. Donenfeld" <Jason@...c4.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: Thomas Gleixner <tglx@...utronix.de>,
LKML <linux-kernel@...r.kernel.org>,
Clemens Ladisch <clemens@...isch.de>,
Sultan Alsawaf <sultan@...neltoast.com>,
Waiman Long <longman@...hat.com>, X86 ML <x86@...nel.org>
Subject: Re: infinite loop in read_hpet from ktime_get_boot_fast_ns
Hi Peter,
On Wed, Jun 12, 2019 at 2:29 PM Peter Zijlstra <peterz@...radead.org> wrote:
> > and the comment at the top mentions explicit sleep hooks. I wasn't
> > sure which function to use from here, though.
>
> Either local_clock() or cpu_clock(cpu). The sleep hooks are not
> something the consumer has to worry about.
I'm not sure whether this is another bug or just a misunderstanding on
what we meant about sleep, but I observed the following behavior,
below, with local_clock(). Notably, local_clock() isn't advanced
during sleep in the way that the other mechanisms are.
Regards,
Jason
[50865.699714] wireguard: local_clock: 1, jiffies: 1,
jiffies_monotoany: 1, ktime_boot: 1, ktime_boot_coarse: 1
[50866.723923] wireguard: local_clock: 2, jiffies: 2,
jiffies_monotoany: 2, ktime_boot: 2, ktime_boot_coarse: 2
[50867.747896] wireguard: local_clock: 3, jiffies: 3,
jiffies_monotoany: 3, ktime_boot: 3, ktime_boot_coarse: 3
[50868.772297] wireguard: local_clock: 4, jiffies: 4,
jiffies_monotoany: 4, ktime_boot: 4, ktime_boot_coarse: 4
[50869.796419] wireguard: local_clock: 5, jiffies: 5,
jiffies_monotoany: 5, ktime_boot: 5, ktime_boot_coarse: 5
[50870.820719] wireguard: local_clock: 6, jiffies: 6,
jiffies_monotoany: 6, ktime_boot: 6, ktime_boot_coarse: 6
[50871.834768] wireguard: local_clock: 7, jiffies: 7,
jiffies_monotoany: 10, ktime_boot: 10, ktime_boot_coarse: 10
[50869.918760] PM: suspend entry (deep)
[50872.846134] PM: suspend exit
[50872.874955] wireguard: local_clock: 8, jiffies: 8,
jiffies_monotoany: 11, ktime_boot: 11, ktime_boot_coarse: 11
[50873.899142] wireguard: local_clock: 9, jiffies: 9,
jiffies_monotoany: 12, ktime_boot: 12, ktime_boot_coarse: 12
[50874.923368] wireguard: local_clock: 10, jiffies: 10,
jiffies_monotoany: 13, ktime_boot: 13, ktime_boot_coarse: 13
[50875.947641] wireguard: local_clock: 11, jiffies: 11,
jiffies_monotoany: 14, ktime_boot: 14, ktime_boot_coarse: 14
[50876.971833] wireguard: local_clock: 12, jiffies: 12,
jiffies_monotoany: 15, ktime_boot: 15, ktime_boot_coarse: 15
[50877.995969] wireguard: local_clock: 13, jiffies: 13,
jiffies_monotoany: 16, ktime_boot: 16, ktime_boot_coarse: 16
[50879.020220] wireguard: local_clock: 14, jiffies: 14,
jiffies_monotoany: 17, ktime_boot: 17, ktime_boot_coarse: 17
[50880.044395] wireguard: local_clock: 15, jiffies: 15,
jiffies_monotoany: 18, ktime_boot: 18, ktime_boot_coarse: 18
static void ugh(struct work_struct *w)
{
int i;
u64 start_localclock = local_clock(), end_localclock;
u64 start_jiffies = jiffies64_to_nsecs(get_jiffies_64()), end_jiffies;
u64 start_jiffiesmonotoany =
ktime_to_ns(ktime_mono_to_any(ns_to_ktime(jiffies64_to_nsecs(get_jiffies_64())),
TK_OFFS_BOOT)), end_jiffiesmonotoany;
u64 start_ktimeboot = ktime_get_boot_ns(), end_ktimeboot;
u64 start_ktimebootcoarse =
ktime_to_ns(ktime_get_coarse_boottime()), end_ktimebootcoarse;
for (i = 0; i < 15; ++i) {
msleep(1000);
end_localclock = local_clock();
end_jiffies = jiffies64_to_nsecs(get_jiffies_64());
end_jiffiesmonotoany =
ktime_to_ns(ktime_mono_to_any(ns_to_ktime(jiffies64_to_nsecs(get_jiffies_64())),
TK_OFFS_BOOT));
end_ktimeboot = ktime_get_boot_ns();
end_ktimebootcoarse = ktime_to_ns(ktime_get_coarse_boottime());
pr_err("local_clock: %llu, jiffies: %llu, jiffies_monotoany: %llu,
ktime_boot: %llu, ktime_boot_coarse: %llu\n",
(end_localclock - start_localclock) / NSEC_PER_SEC,
(end_jiffies - start_jiffies) / NSEC_PER_SEC,
(end_jiffiesmonotoany - start_jiffiesmonotoany) / NSEC_PER_SEC,
(end_ktimeboot - start_ktimeboot) / NSEC_PER_SEC,
(end_ktimebootcoarse - start_ktimebootcoarse) / NSEC_PER_SEC);
}
}
static DECLARE_WORK(blah, ugh);
static int __init mod_init(void)
{
schedule_work(&blah);
return 0;
}
Powered by blists - more mailing lists