From: "Sverdlin, Alexander (Nokia - DE/Ulm)" <alexander.sverdlin@nokia.com>
To: "x86@kernel.org" <x86@kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Cc: "Sverdlin,
Alexander (Nokia - DE/Ulm)" <alexander.sverdlin@nokia.com>,
Andy Lutomirski <luto@kernel.org>, Ingo Molnar <mingo@redhat.com>,
Borislav Petkov <bp@alien8.de>, "H. Peter Anvin" <hpa@zytor.com>,
Thomas Gleixner <tglx@linutronix.de>,
Jason Vas Dias <jason.vas.dias@gmail.com>
Subject: [PATCH v2] x86/vdso: implement clock_gettime(CLOCK_MONOTONIC_RAW, ...)
Date: Tue, 4 Jun 2019 16:42:20 +0000 [thread overview]
Message-ID: <20190604164117.22154-1-alexander.sverdlin@nokia.com> (raw)
In-Reply-To: <alpine.DEB.2.21.1905281055240.1859@nanos.tec.linutronix.de>
From: Alexander Sverdlin <alexander.sverdlin@nokia.com>
Add CLOCK_MONOTONIC_RAW to the existing clock_gettime() vDSO
implementation. This is based on the ideas of Jason Vas Dias and comments
of Thomas Gleixner.
---- Test code ----
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <unistd.h>
#define CLOCK_TYPE CLOCK_MONOTONIC_RAW
#define DURATION_SEC 10
int main(int argc, char **argv)
{
struct timespec t, end;
unsigned long long cnt = 0;
clock_gettime(CLOCK_TYPE, &end);
end.tv_sec += DURATION_SEC;
do {
clock_gettime(CLOCK_TYPE, &t);
++cnt;
} while (t.tv_sec < end.tv_sec || t.tv_nsec < end.tv_nsec);
dprintf(STDOUT_FILENO, "%llu", cnt);
return EXIT_SUCCESS;
}
-------------------
The results from the above test program:
Clock Before After Diff
----- ------ ----- ----
CLOCK_MONOTONIC 355.5M 355.5M
CLOCK_MONOTONIC_RAW 44.9M 371.2M +726%
CLOCK_REALTIME 355.5M 355.5M
Link: https://lore.kernel.org/patchwork/patch/933583/
Link: https://bugzilla.kernel.org/show_bug.cgi?id=198961
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jason Vas Dias <jason.vas.dias@gmail.com>
Signed-off-by: Alexander Sverdlin <alexander.sverdlin@nokia.com>
---
Changelog:
v2: copy do_hres() into do_monotonic_raw()
arch/x86/entry/vdso/vclock_gettime.c | 35 +++++++++++++++++++++++++++++++++
arch/x86/entry/vsyscall/vsyscall_gtod.c | 6 ++++++
arch/x86/include/asm/vgtod.h | 2 ++
3 files changed, 43 insertions(+)
diff --git a/arch/x86/entry/vdso/vclock_gettime.c b/arch/x86/entry/vdso/vclock_gettime.c
index 0f82a70..64736a4 100644
--- a/arch/x86/entry/vdso/vclock_gettime.c
+++ b/arch/x86/entry/vdso/vclock_gettime.c
@@ -168,6 +168,39 @@ notrace static int do_hres(clockid_t clk, struct timespec *ts)
return 0;
}
+/*
+ * Attempts to merge the below copy with the above routine led to 5% performance
+ * drop (CLOCK_MONOTONIC, CLOCK_REALTIME) up to now. Test before making changes.
+ */
+notrace static int do_monotonic_raw(struct timespec *ts)
+{
+ struct vgtod_ts *base = >od->basetime[CLOCK_MONOTONIC_RAW];
+ u64 cycles, last, sec, ns;
+ unsigned int seq;
+
+ do {
+ seq = gtod_read_begin(gtod);
+ cycles = vgetcyc(gtod->vclock_mode);
+ ns = base->nsec;
+ last = gtod->cycle_last;
+ if (unlikely((s64)cycles < 0))
+ return vdso_fallback_gettime(CLOCK_MONOTONIC_RAW, ts);
+ if (cycles > last)
+ ns += (cycles - last) * gtod->raw_mult;
+ ns >>= gtod->raw_shift;
+ sec = base->sec;
+ } while (unlikely(gtod_read_retry(gtod, seq)));
+
+ /*
+ * Do this outside the loop: a race inside the loop could result
+ * in __iter_div_u64_rem() being extremely slow.
+ */
+ ts->tv_sec = sec + __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+ ts->tv_nsec = ns;
+
+ return 0;
+}
+
notrace static void do_coarse(clockid_t clk, struct timespec *ts)
{
struct vgtod_ts *base = >od->basetime[clk];
@@ -199,6 +232,8 @@ notrace int __vdso_clock_gettime(clockid_t clock, struct timespec *ts)
do_coarse(clock, ts);
return 0;
}
+ if (clock == CLOCK_MONOTONIC_RAW)
+ return do_monotonic_raw(ts);
return vdso_fallback_gettime(clock, ts);
}
diff --git a/arch/x86/entry/vsyscall/vsyscall_gtod.c b/arch/x86/entry/vsyscall/vsyscall_gtod.c
index cfcdba0..9f7744f3 100644
--- a/arch/x86/entry/vsyscall/vsyscall_gtod.c
+++ b/arch/x86/entry/vsyscall/vsyscall_gtod.c
@@ -46,6 +46,8 @@ void update_vsyscall(struct timekeeper *tk)
vdata->mask = tk->tkr_mono.mask;
vdata->mult = tk->tkr_mono.mult;
vdata->shift = tk->tkr_mono.shift;
+ vdata->raw_mult = tk->tkr_raw.mult;
+ vdata->raw_shift = tk->tkr_raw.shift;
base = &vdata->basetime[CLOCK_REALTIME];
base->sec = tk->xtime_sec;
@@ -65,6 +67,10 @@ void update_vsyscall(struct timekeeper *tk)
}
base->nsec = nsec;
+ base = &vdata->basetime[CLOCK_MONOTONIC_RAW];
+ base->sec = tk->raw_sec;
+ base->nsec = tk->tkr_raw.xtime_nsec;
+
base = &vdata->basetime[CLOCK_REALTIME_COARSE];
base->sec = tk->xtime_sec;
base->nsec = tk->tkr_mono.xtime_nsec >> tk->tkr_mono.shift;
diff --git a/arch/x86/include/asm/vgtod.h b/arch/x86/include/asm/vgtod.h
index 913a133..65ac320 100644
--- a/arch/x86/include/asm/vgtod.h
+++ b/arch/x86/include/asm/vgtod.h
@@ -43,6 +43,8 @@ struct vsyscall_gtod_data {
u64 mask;
u32 mult;
u32 shift;
+ u32 raw_mult;
+ u32 raw_shift;
struct vgtod_ts basetime[VGTOD_BASES];
--
2.4.6
next prev parent reply other threads:[~2019-06-04 16:42 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-05-28 8:01 [PATCH] x86/vdso: implement clock_gettime(CLOCK_MONOTONIC_RAW, ...) Sverdlin, Alexander (Nokia - DE/Ulm)
2019-05-28 18:01 ` Thomas Gleixner
2019-06-04 16:42 ` Sverdlin, Alexander (Nokia - DE/Ulm) [this message]
2019-06-04 17:39 ` [PATCH v2] " Andy Lutomirski
2019-06-05 14:41 ` [PATCH v3 0/2] x86/vdso: CLOCK_MONOTONIC_RAW implementation Sverdlin, Alexander (Nokia - DE/Ulm)
2019-06-05 14:41 ` [PATCH v3 1/2] x86/vdso: Move mult and shift into struct vgtod_ts Sverdlin, Alexander (Nokia - DE/Ulm)
2019-06-23 10:18 ` Thomas Gleixner
2019-06-24 9:16 ` Sverdlin, Alexander (Nokia - DE/Ulm)
2019-06-24 9:40 ` Thomas Gleixner
2019-06-27 12:00 ` Sverdlin, Alexander (Nokia - DE/Ulm)
2019-06-27 12:07 ` Thomas Gleixner
2019-06-27 12:15 ` Sverdlin, Alexander (Nokia - DE/Ulm)
2019-06-27 12:21 ` Thomas Gleixner
2019-06-05 14:41 ` [PATCH v3 2/2] x86/vdso: implement clock_gettime(CLOCK_MONOTONIC_RAW, ...) Sverdlin, Alexander (Nokia - DE/Ulm)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190604164117.22154-1-alexander.sverdlin@nokia.com \
--to=alexander.sverdlin@nokia.com \
--cc=bp@alien8.de \
--cc=hpa@zytor.com \
--cc=jason.vas.dias@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=luto@kernel.org \
--cc=mingo@redhat.com \
--cc=tglx@linutronix.de \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).