From: Vincenzo Frascino <vincenzo.frascino@arm.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-arch@vger.kernel.org,
LAK <linux-arm-kernel@lists.infradead.org>,
LKML <linux-kernel@vger.kernel.org>,
linux-mips@vger.kernel.org, linux-kselftest@vger.kernel.org,
catalin.marinas@arm.com, Will Deacon <will.deacon@arm.com>,
Arnd Bergmann <arnd@arndb.de>,
linux@armlinux.org.uk, Ralf Baechle <ralf@linux-mips.org>,
paul.burton@mips.com, Daniel Lezcano <daniel.lezcano@linaro.org>,
salyzyn@android.com, pcc@google.com, shuah@kernel.org,
0x7f454c46@gmail.com, linux@rasmusvillemoes.dk,
huw@codeweavers.com, sthotton@marvell.com,
andre.przywara@arm.com, Andy Lutomirski <luto@kernel.org>
Subject: Re: lib/vdso: Make delta calculation work correctly
Date: Wed, 26 Jun 2019 12:08:29 +0100 [thread overview]
Message-ID: <601481b9-aced-d03c-03a3-3de36bc58ffb@arm.com> (raw)
In-Reply-To: <alpine.DEB.2.21.1906261159230.32342@nanos.tec.linutronix.de>
Hi Thomas,
On 26/06/2019 11:02, Thomas Gleixner wrote:
> The x86 vdso implementation on which the generic vdso library is based on
> has subtle (unfortunately undocumented) twists:
>
> 1) The code assumes that the clocksource mask is U64_MAX which means that
> no bits are masked. Which is true for any valid x86 VDSO clocksource.
> Stupidly it still did the mask operation for no reason and at the wrong
> place right after reading the clocksource.
>
> 2) It contains a sanity check to catch the case where slightly
> unsynchronized TSC values can be overserved which would cause the delta
> calculation to make a huge jump. It therefore checks whether the
> current TSC value is larger than the value on which the current
> conversion is based on. If it's not larger the base value is used to
> prevent time jumps.
>
> #1 Is not only stupid for the X86 case because it does the masking for no
> reason it is also completely wrong for clocksources with a smaller mask
> which can legitimately wrap around during a conversion period. The core
> timekeeping code does it correct by applying the mask after the delta
> calculation:
>
> (now - base) & mask
>
> #2 is equally broken for clocksources which have smaller masks and can wrap
> around during a conversion period because there the now > base check is
> just wrong and causes stale time stamps and time going backwards issues.
>
> Unbreak it by:
>
> 1) Removing the mask operation from the clocksource read which makes the
> fallback detection work for all clocksources
>
> 2) Replacing the conditional delta calculation with a overrideable inline
> function.
>
> #2 could reuse clocksource_delta() from the timekeeping code but that
> results in a significant performance hit for the x86 VSDO. The timekeeping
> core code must have the non optimized version as it has to operate
> correctly with clocksources which have smaller masks as well to handle the
> case where TSC is discarded as timekeeper clocksource and replaced by HPET
> or pmtimer. For the VDSO there is no replacement clocksource. If TSC is
> unusable the syscall is enforced which does the right thing.
>
> To accomodate to the needs of various architectures provide an overrideable
> inline function which defaults to the regular delta calculation with
> masking:
>
> (now - base) & mask
>
> Override it for x86 with the non-masking and checking version.
>
> This unbreaks the ARM64 syscall fallback operation, allows to use
> clocksources with arbitrary width and preserves the performance
> optimization for x86.
>
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
A part a typo that leads to compilation errors on non-x86 platforms the rest
looks fine by me.
I tested it on arm64 and behaves correctly.
With this:
Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
> ---
> arch/x86/include/asm/vdso/gettimeofday.h | 27 +++++++++++++++++++++++++++
> lib/vdso/gettimeofday.c | 19 +++++++++++++++----
> 2 files changed, 42 insertions(+), 4 deletions(-)
>
> --- a/arch/x86/include/asm/vdso/gettimeofday.h
> +++ b/arch/x86/include/asm/vdso/gettimeofday.h
> @@ -229,6 +229,33 @@ static __always_inline const struct vdso
> return __vdso_data;
> }
>
> +/*
> + * x86 specific delta calculation.
> + *
> + * The regular implementation assumes that clocksource reads are globally
> + * monotonic. The TSC can be slightly off across sockets which can cause
> + * the regular delta calculation (@cycles - @last) to return a huge time
> + * jump.
> + *
> + * Therefore it needs to be verified that @cycles are greater than
> + * @last. If not then use @last, which is the base time of the current
> + * conversion period.
> + *
> + * This variant also removes the masking of the subtraction because the
> + * clocksource mask of all VDSO capable clocksources on x86 is U64_MAX
> + * which would result in a pointless operation. The compiler cannot
> + * optimize it away as the mask comes from the vdso data and is not compile
> + * time constant.
> + */
> +static __always_inline
> +u64 vdso_calc_delta(u64 cycles, u64 last, u64 mask, u32 mult)
> +{
> + if (cycles > last)
> + return (cycles - last) * mult;
> + return 0;
> +}
> +#define vdso_calc_delta vdso_calc_delta
> +
> #endif /* !__ASSEMBLY__ */
>
> #endif /* __ASM_VDSO_GETTIMEOFDAY_H */
> --- a/lib/vdso/gettimeofday.c
> +++ b/lib/vdso/gettimeofday.c
> @@ -26,6 +26,18 @@
> #include <asm/vdso/gettimeofday.h>
> #endif /* ENABLE_COMPAT_VDSO */
>
> +#ifndef vdso_calc_delta
> +/*
> + * Default implementation which works for all sane clocksources. That
> + * obviously excludes x86/TSC.
> + */
> +static __always_inline
> +u64 vdso_calc_delta(u64 cycles, u64 last, u64 mask, u32 mult)
> +{
> + return ((cyles - last) & mask) * mult;
Typo here:
s/cyles/cycles/
> +}
> +#endif
> +
> static int do_hres(const struct vdso_data *vd, clockid_t clk,
> struct __kernel_timespec *ts)
> {
> @@ -35,14 +47,13 @@ static int do_hres(const struct vdso_dat
>
> do {
> seq = vdso_read_begin(vd);
> - cycles = __arch_get_hw_counter(vd->clock_mode) &
> - vd->mask;
> + cycles = __arch_get_hw_counter(vd->clock_mode);
> ns = vdso_ts->nsec;
> last = vd->cycle_last;
> if (unlikely((s64)cycles < 0))
> return clock_gettime_fallback(clk, ts);
> - if (cycles > last)
> - ns += (cycles - last) * vd->mult;
> +
> + ns += vdso_calc_delta(cycles, last, vd->mask, vd->mult);
> ns >>= vd->shift;
> sec = vdso_ts->sec;
> } while (unlikely(vdso_read_retry(vd, seq)));
>
--
Regards,
Vincenzo
next prev parent reply other threads:[~2019-06-26 11:08 UTC|newest]
Thread overview: 108+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-06-21 9:52 [PATCH v7 00/25] Unify vDSOs across more architectures Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 01/25] kernel: Standardize vdso_datapage Vincenzo Frascino
2019-06-24 13:56 ` Catalin Marinas
2019-06-21 9:52 ` [PATCH v7 02/25] kernel: Define gettimeofday vdso common code Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 03/25] kernel: Unify update_vsyscall implementation Vincenzo Frascino
2019-06-21 10:49 ` Huw Davies
2019-06-21 9:52 ` [PATCH v7 04/25] arm64: Substitute gettimeofday with C implementation Vincenzo Frascino
2019-06-24 13:36 ` Will Deacon
2019-06-24 13:59 ` Vincenzo Frascino
2019-06-25 16:18 ` [PATCH 1/3] lib/vdso: Delay mask application in do_hres() Vincenzo Frascino
2019-06-25 16:18 ` [PATCH 2/3] arm64: Fix __arch_get_hw_counter() implementation Vincenzo Frascino
2019-06-25 16:18 ` [PATCH 3/3] arm64: compat: " Vincenzo Frascino
2019-06-25 17:02 ` [PATCH 1/3] lib/vdso: Delay mask application in do_hres() Thomas Gleixner
2019-06-25 18:27 ` Thomas Gleixner
2019-06-25 20:15 ` Andy Lutomirski
2019-06-25 22:24 ` Thomas Gleixner
2019-06-26 6:38 ` Thomas Gleixner
2019-06-26 9:25 ` Vincenzo Frascino
2019-06-26 10:02 ` lib/vdso: Make delta calculation work correctly Thomas Gleixner
2019-06-26 11:08 ` Vincenzo Frascino [this message]
2019-06-24 13:58 ` [PATCH v7 04/25] arm64: Substitute gettimeofday with C implementation Catalin Marinas
2019-06-25 15:33 ` Dave Martin
2019-06-26 13:27 ` Vincenzo Frascino
2019-06-26 16:14 ` Dave Martin
2019-06-26 19:01 ` Vincenzo Frascino
2019-06-27 10:01 ` Dave Martin
2019-06-27 10:57 ` Vincenzo Frascino
2019-06-27 11:27 ` Dave Martin
2019-06-27 11:59 ` Vincenzo Frascino
2019-06-27 14:38 ` Dave Martin
2019-06-27 15:34 ` Vincenzo Frascino
2019-06-25 17:43 ` [PATCH] arm64: vdso: Fix compilation with clang < 8 Vincenzo Frascino
2019-06-26 11:36 ` [PATCH v2] arm64: vdso: Fix compilation with clang older then 8 Vincenzo Frascino
[not found] ` <CGME20190628130921eucas1p239935b0771032c331911eacc1a69dd2e@eucas1p2.samsung.com>
2019-06-28 13:09 ` [PATCH v7 04/25] arm64: Substitute gettimeofday with C implementation Marek Szyprowski
2019-06-28 14:32 ` Vincenzo Frascino
2019-06-28 16:50 ` Sylwester Nawrocki
2019-06-29 6:58 ` Vincenzo Frascino
2019-07-08 12:57 ` Sylwester Nawrocki
2019-07-08 13:09 ` Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 05/25] arm64: Build vDSO with -ffixed-x18 Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 06/25] arm64: compat: Add missing syscall numbers Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 07/25] arm64: compat: Expose signal related structures Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 08/25] arm64: compat: Generate asm offsets for signals Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 09/25] lib: vdso: Add compat support Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 10/25] arm64: compat: Add vDSO Vincenzo Frascino
2019-06-24 14:00 ` Catalin Marinas
2019-07-10 4:02 ` John Stultz
2019-07-10 6:12 ` Thomas Gleixner
2019-07-10 9:48 ` Vincenzo Frascino
2019-07-10 8:27 ` Will Deacon
2019-07-10 8:58 ` Thomas Gleixner
2019-07-10 9:12 ` Will Deacon
2019-07-10 9:47 ` Vincenzo Frascino
2019-07-10 13:41 ` Vincenzo Frascino
2019-07-10 13:04 ` [PATCH] arm64: vdso: Fix ABI regression in compat vdso Vincenzo Frascino
2019-07-10 13:25 ` Will Deacon
2019-07-10 13:42 ` Vincenzo Frascino
2019-07-10 14:01 ` [PATCH v2] " Vincenzo Frascino
2019-07-10 15:44 ` John Stultz
2019-07-10 15:53 ` Vincenzo Frascino
2019-07-11 9:45 ` Will Deacon
2019-07-11 10:34 ` Thomas Gleixner
2019-07-11 11:32 ` Will Deacon
2019-06-21 9:52 ` [PATCH v7 11/25] arm64: Refactor vDSO code Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 12/25] arm64: compat: vDSO setup for compat layer Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 13/25] arm64: elf: vDSO code page discovery Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 14/25] arm64: compat: Get sigreturn trampolines from vDSO Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 15/25] arm64: Add vDSO compat support Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 16/25] arm: Add support for generic vDSO Vincenzo Frascino
2019-12-04 13:51 ` [PATCH v7 16/25] arm: Add support for generic vDSO (causing crash) Guenter Roeck
2019-12-04 13:58 ` Vincenzo Frascino
2019-12-04 16:16 ` Guenter Roeck
2019-12-04 17:15 ` Vincenzo Frascino
2019-12-04 19:39 ` Guenter Roeck
2019-12-05 9:42 ` Philippe Mathieu-Daudé
2019-12-05 10:00 ` Vincenzo Frascino
2019-12-05 11:02 ` Arnd Bergmann
2019-12-05 14:56 ` Philippe Mathieu-Daudé
2019-06-21 9:52 ` [PATCH v7 17/25] arm: Add clock_getres entry point Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 18/25] arm: Add clock_gettime64 " Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 19/25] mips: Add support for generic vDSO Vincenzo Frascino
2019-07-26 5:15 ` Paul Burton
2019-07-26 16:29 ` [PATCH 0/2] mips: vdso: Fix Makefile Vincenzo Frascino
2019-07-26 16:29 ` [PATCH 1/2] mips: vdso: Fix source path Vincenzo Frascino
2019-07-26 16:29 ` [PATCH 2/2] mips: vdso: Fix flip/flop vdso building bug Vincenzo Frascino
2019-07-28 22:20 ` [PATCH 0/2] mips: vdso: Fix Makefile Paul Burton
2019-06-21 9:52 ` [PATCH v7 20/25] mips: Add clock_getres entry point Vincenzo Frascino
2019-07-26 5:15 ` Paul Burton
2019-06-21 9:52 ` [PATCH v7 21/25] mips: Add clock_gettime64 " Vincenzo Frascino
2019-07-26 5:15 ` Paul Burton
2019-06-21 9:52 ` [PATCH v7 22/25] x86: Add support for generic vDSO Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 23/25] x86: Add clock_getres entry point Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 24/25] x86: Add clock_gettime64 " Vincenzo Frascino
2019-06-21 9:52 ` [PATCH v7 25/25] kselftest: Extend vDSO selftest Vincenzo Frascino
2019-06-24 0:34 ` [PATCH v7 00/25] Unify vDSOs across more architectures Thomas Gleixner
2019-06-24 1:15 ` Andy Lutomirski
2019-06-24 7:42 ` Thomas Gleixner
2019-06-24 13:21 ` Vincenzo Frascino
2019-06-24 14:18 ` Thomas Gleixner
2019-06-24 14:23 ` Russell King - ARM Linux admin
2019-06-24 14:49 ` Catalin Marinas
2019-06-24 16:20 ` Vincenzo Frascino
2019-10-25 11:42 ` Geert Uytterhoeven
2019-06-24 18:41 ` Paul Burton
2019-06-24 23:16 ` Vincenzo Frascino
2019-06-25 17:11 ` Paul Burton
2019-06-25 17:17 ` Vincenzo Frascino
2019-06-24 12:50 ` Andre Przywara
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=601481b9-aced-d03c-03a3-3de36bc58ffb@arm.com \
--to=vincenzo.frascino@arm.com \
--cc=0x7f454c46@gmail.com \
--cc=andre.przywara@arm.com \
--cc=arnd@arndb.de \
--cc=catalin.marinas@arm.com \
--cc=daniel.lezcano@linaro.org \
--cc=huw@codeweavers.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-kselftest@vger.kernel.org \
--cc=linux-mips@vger.kernel.org \
--cc=linux@armlinux.org.uk \
--cc=linux@rasmusvillemoes.dk \
--cc=luto@kernel.org \
--cc=paul.burton@mips.com \
--cc=pcc@google.com \
--cc=ralf@linux-mips.org \
--cc=salyzyn@android.com \
--cc=shuah@kernel.org \
--cc=sthotton@marvell.com \
--cc=tglx@linutronix.de \
--cc=will.deacon@arm.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;
as well as URLs for NNTP newsgroup(s).