From: Zhan Xusheng <zhanxusheng1024@gmail.com>
To: tglx@kernel.org, luto@kernel.org, vincenzo.frascino@arm.com
Cc: thomas.weissschuh@linutronix.de, david.laight.linux@gmail.com,
linux-kernel@vger.kernel.org, zhanxusheng@xiaomi.com
Subject: [PATCH v3 0/3] vdso: Keep the CLOCK_AUX base at full precision
Date: Tue, 1 Sep 2026 10:06:33 +0800 [thread overview]
Message-ID: <20260901020636.1821993-1-zhanxusheng@xiaomi.com> (raw)
v1: https://lore.kernel.org/all/20260831125557.1490398-1-zhanxusheng@xiaomi.com
v2 went out before David's mail arrived. It touched only the call sites,
so both of his remarks about the helper applied to it unchanged.
Changes in v3:
- The quotient is a u32, as David suggested and as the u32-divisor
version already returns. vsyscall.o loses 16 bytes of text on x86-64
and the loop drops from 36 to 28 instructions on 32-bit gcc.
- The barrier keeps the value in a register instead of offering a memory
alternative. David asked whether clang makes a mess of "+rm": it does,
it takes the memory alternative and spills inside the loop, on 64-bit
as well. Both forms still stop the loop becoming a division, checked
for gcc and clang at both widths.
- Fixes: tag on 2/3.
Changes in v2, from Thomas' review of 1/3:
- Drop the nsec_per_sec local and spell the divisor out at each site.
- Store the remainder straight into the basetime, as the coarse clocks
already do. That also makes the copy of the CLOCK_MONOTONIC values for
CLOCK_BOOTTIME take sec and nsec from the same place rather than one
from the basetime and one from a local left over from the division.
Not a codegen win: gcc emits the same instructions either way.
- Keep "vdso_ts->sec = sec;" on its own line under the copy comment.
- Picked up the Reviewed-by on 3/3, which is unchanged.
__iter_div_u64_rem() has the same "+rm" barrier, and it sits on the
clock_gettime() path through vdso_set_timespec(). Converting it is worth
64 bytes of vdso64 text and 112 of vdso32 built with clang 18; gcc is
indifferent. It is not in this series because it also has users in
arch/x86/kvm, timespec64_add_ns() and lib/math/div64.c. Say the word and
I will send it.
The problem 2/3 fixes: the CLOCK_AUX basetime is shifted down to
nanoseconds and back up, which drops the fractional nanoseconds of
xtime_nsec, so the vDSO floors the base and the cycle delta separately
where ktime_get_aux() floors their sum. The vDSO reading ends up 0 or
1 ns below the syscall for the same clock, and clock_getres() advertises
1 ns for these clocks.
Checked under QEMU with an auxiliary clock enabled through
/sys/kernel/time/aux_clocks/0/aux_clock_enable, comparing CLOCK_AUX via
the vDSO against the raw syscall at offset 0, at +5.123456789 s, and with
offs_aux driven negative. The two agree within read latency in all three,
and over 100000 interleaved pairs each the vDSO reading is never ahead of a
syscall reading taken after it. The 1 ns bias itself is not measurable
that way, a pair of reads costing several hundred nanoseconds. The
argument for it is the algebra in 2/3, plus a sweep over
(xtime_nsec, delta) which puts the 1 ns case at 60% for shift 24.
Zhan Xusheng (3):
vdso/math64: Add and use __iter_div64_u64_rem()
vdso/vsyscall: Keep the CLOCK_AUX base scaled
vdso/gettimeofday: Assert that the clock id fits the dispatch mask
include/vdso/math64.h | 22 ++++++++++++++++++++++
kernel/time/vsyscall.c | 28 ++++++++++++----------------
lib/vdso/gettimeofday.c | 2 ++
3 files changed, 36 insertions(+), 16 deletions(-)
base-commit: abdf623ddb75b24659018d3952d8f61937306ae5
--
2.43.0
next reply other threads:[~2026-09-01 2:06 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-01 2:06 Zhan Xusheng [this message]
2026-09-01 2:06 ` [PATCH v3 1/3] vdso/math64: Add and use __iter_div64_u64_rem() Zhan Xusheng
2026-09-01 8:35 ` Thomas Weißschuh
2026-09-01 12:48 ` David Laight
2026-09-01 13:05 ` Thomas Weißschuh
2026-09-01 2:06 ` [PATCH v3 2/3] vdso/vsyscall: Keep the CLOCK_AUX base scaled Zhan Xusheng
2026-09-01 8:42 ` Thomas Weißschuh
2026-09-01 2:06 ` [PATCH v3 3/3] vdso/gettimeofday: Assert that the clock id fits the dispatch mask Zhan Xusheng
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=20260901020636.1821993-1-zhanxusheng@xiaomi.com \
--to=zhanxusheng1024@gmail.com \
--cc=david.laight.linux@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=luto@kernel.org \
--cc=tglx@kernel.org \
--cc=thomas.weissschuh@linutronix.de \
--cc=vincenzo.frascino@arm.com \
--cc=zhanxusheng@xiaomi.com \
/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
all inboxes | Powered by JetHome®