Received-SPF: pass (google.com: domain of linux-kernel+bounces-23152-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Wed, 10 Jan 2024 23:31:00 -0800
From: Richard Cochran <richardcochran@gmail.com>
To: Arnd Bergmann <arnd@arndb.de>
Cc: Sagi Maimon <maimon.sagi@gmail.com>, Andy Lutomirski <luto@kernel.org>,
	tglx@linutronix.de, Ingo Molnar <mingo@redhat.com>,
	Borislav Petkov <bp@alien8.de>,
	Dave Hansen <dave.hansen@linux.intel.com>, x86@kernel.org,
	"H. Peter Anvin" <hpa@zytor.com>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Sohil Mehta <sohil.mehta@intel.com>,
	Rick Edgecombe <rick.p.edgecombe@intel.com>,
	Nhat Pham <nphamcs@gmail.com>, Palmer Dabbelt <palmer@sifive.com>,
	Kees Cook <keescook@chromium.org>,
	Alexey Gladkov <legion@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>, linux-kernel@vger.kernel.org,
	linux-api@vger.kernel.org, Linux-Arch <linux-arch@vger.kernel.org>,
	Netdev <netdev@vger.kernel.org>
Subject: Re: [PATCH v3] posix-timers: add multi_clock_gettime system call
Message-ID: <ZZ-ZNHgDsZwg9CaW@hoboy.vegasvil.org>
References: <20231228122411.3189-1-maimon.sagi@gmail.com>
 <f254c189-463e-43a3-bc09-9a8869ebf819@app.fastmail.com>
 <CAMuE1bF0Hho4VwO6w3f+9z3j5TtscYzuAjj10MFt2mZXG2P8dQ@mail.gmail.com>
 <84d8e9d7-09ce-4781-8dfa-a74bb0955ae8@app.fastmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <84d8e9d7-09ce-4781-8dfa-a74bb0955ae8@app.fastmail.com>

On Tue, Jan 02, 2024 at 12:29:59PM +0100, Arnd Bergmann wrote:

> I think Andy's suggestion of exposing time offsets instead
> of absolute times would actually achieve that: If the
> interface is changed to return the offset against
> CLOCK_MONOTONIC, CLOCK_MONOTONIC_RAW or CLOCK_BOOTTIME
> (not sure what is best here), then the new syscall can use
> getcrosststamp() where supported for the best results or
> fall back to gettimex64() or gettime64() otherwise to
> provide a consistent user interface.

Yes, it makes more sense to provide the offset, since that is what the
user needs in the end.

Can we change the name of the system call to "clock compare"?

int clock_compare(clockid_t a, clockid_t b,
		  int64_t *offset, int64_t *error);

returns: zero or error code,
 offset = a - b
 error  = maximum error due to asymmetry

If clocks a and b are both System-V clocks, then *error=0 and *offset
can be returned directly from the kernel's time keeping state.

If getcrosststamp() is supported on a or b, then invoke it.

otherwise do this:

   t1 = gettime(a)
   t2 = gettime(b)
   t3 - gettime(c)

   *offset = (t1 + t3)/2 - t2
   *error  = (t3 - t1)/2

There is no need for repeated measurement, since user space can call
again when `error` is unacceptable.

Thanks,
Richard