From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751816AbdFGViR (ORCPT ); Wed, 7 Jun 2017 17:38:17 -0400 Received: from mail-wm0-f51.google.com ([74.125.82.51]:33872 "EHLO mail-wm0-f51.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751779AbdFGViP (ORCPT ); Wed, 7 Jun 2017 17:38:15 -0400 Date: Wed, 7 Jun 2017 23:38:10 +0200 From: Daniel Lezcano To: Alexandre Belloni Cc: Nicolas Ferre , Boris Brezillon , linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Thomas Gleixner Subject: Re: [PATCH 46/58] clocksource/drivers: Add a new driver for the Atmel ARM TC blocks Message-ID: <20170607213810.GK2345@mai> References: <20170530215139.9983-1-alexandre.belloni@free-electrons.com> <20170530215139.9983-47-alexandre.belloni@free-electrons.com> <20170606152104.GC2345@mai> <20170606180559.pkrr7ux2qqnmsd6y@piout.net> <20170607141735.GH2345@mai> <20170607150908.kytgtzwgjjnxtsp3@piout.net> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <20170607150908.kytgtzwgjjnxtsp3@piout.net> User-Agent: Mutt/1.5.24 (2015-08-30) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jun 07, 2017 at 05:09:08PM +0200, Alexandre Belloni wrote: > On 07/06/2017 at 16:17:35 +0200, Daniel Lezcano wrote: > > > > > This driver uses regmap and syscon to be able to probe early in the boot > > > > > and avoid having to switch on the TCB clocksource later. Using regmap also > > > > > means that unused TCB channels may be used by other drivers (PWM for > > > > > example). > > > > > > > > Can you give more details, I fail to understand how regmap and syscon help to > > > > probe sooner than timer_init()? > > > > > > > > > Because before that, the tcb driver relied on atmel_tclib to share the > > > TCBs and it happened way too late, at arch_initcall() time. > > > > So is it still necesary to use regmap? I would like to take the opportunity to > > move the init routine to the common init routine if possible: > > > > https://patchwork.kernel.org/patch/9768845/ > > > > It is still necessary because we want to be able to share the timer > between multiple drivers. For example, you can have the clocksource on > channel 0, clockevent on channel 1 and a pwm on channel 2 The hardware timer can be shared, the channels used in different subsystem. Each channel are used exclusively. What is the benefit of regmap? It has a cost, and takes a lock at each read. For instance: +static u64 tc_get_cycles(struct clocksource *cs) +{ + u32 lower, upper, tmp; + + do { + regmap_read(tc.regmap, ATMEL_TC_CV(1), &upper); + regmap_read(tc.regmap, ATMEL_TC_CV(0), &lower); + regmap_read(tc.regmap, ATMEL_TC_CV(1), &tmp); + } while (upper != tmp); + + return (upper << 16) | lower; +} Is: +static u64 tc_get_cycles(struct clocksource *cs) +{ + u32 lower, upper, tmp; + + do { + regmap_read(tc.regmap, ATMEL_TC_CV(1), &upper); lock(); lot-of-things(); unlock(); + regmap_read(tc.regmap, ATMEL_TC_CV(0), &lower); lock(); lot-of-things(); unlock(); + regmap_read(tc.regmap, ATMEL_TC_CV(1), &tmp); lock(); lot-of-things(); unlock(); + } while (upper != tmp); + + return (upper << 16) | lower; +} I suggest to look what is in 'lot-of-things()' and especially what is doing regcache_read(). May be you can reconsider the regmap? This driver is the only one use the regmap AFAICT and I don't think it is adequate. > > > > Can you explain why we have two clocks here? > > > > > > > > > > Each channel have its clock, I can add a comment if you want. > > > > I don't understand. Why do we have two clocks? > > > > One channel is driven by one clock and the second one takes the overflow signal > > from the first one, so no second clock is involved there, no? > > > > Those are the peripheral clocks, they are not used by the counters but > used to be able to read/write the registers. Mmh, strange. Why is the clk[0]'s rate used in this case? -- Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog From mboxrd@z Thu Jan 1 00:00:00 1970 From: daniel.lezcano@linaro.org (Daniel Lezcano) Date: Wed, 7 Jun 2017 23:38:10 +0200 Subject: [PATCH 46/58] clocksource/drivers: Add a new driver for the Atmel ARM TC blocks In-Reply-To: <20170607150908.kytgtzwgjjnxtsp3@piout.net> References: <20170530215139.9983-1-alexandre.belloni@free-electrons.com> <20170530215139.9983-47-alexandre.belloni@free-electrons.com> <20170606152104.GC2345@mai> <20170606180559.pkrr7ux2qqnmsd6y@piout.net> <20170607141735.GH2345@mai> <20170607150908.kytgtzwgjjnxtsp3@piout.net> Message-ID: <20170607213810.GK2345@mai> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Wed, Jun 07, 2017 at 05:09:08PM +0200, Alexandre Belloni wrote: > On 07/06/2017 at 16:17:35 +0200, Daniel Lezcano wrote: > > > > > This driver uses regmap and syscon to be able to probe early in the boot > > > > > and avoid having to switch on the TCB clocksource later. Using regmap also > > > > > means that unused TCB channels may be used by other drivers (PWM for > > > > > example). > > > > > > > > Can you give more details, I fail to understand how regmap and syscon help to > > > > probe sooner than timer_init()? > > > > > > > > > Because before that, the tcb driver relied on atmel_tclib to share the > > > TCBs and it happened way too late, at arch_initcall() time. > > > > So is it still necesary to use regmap? I would like to take the opportunity to > > move the init routine to the common init routine if possible: > > > > https://patchwork.kernel.org/patch/9768845/ > > > > It is still necessary because we want to be able to share the timer > between multiple drivers. For example, you can have the clocksource on > channel 0, clockevent on channel 1 and a pwm on channel 2 The hardware timer can be shared, the channels used in different subsystem. Each channel are used exclusively. What is the benefit of regmap? It has a cost, and takes a lock at each read. For instance: +static u64 tc_get_cycles(struct clocksource *cs) +{ + u32 lower, upper, tmp; + + do { + regmap_read(tc.regmap, ATMEL_TC_CV(1), &upper); + regmap_read(tc.regmap, ATMEL_TC_CV(0), &lower); + regmap_read(tc.regmap, ATMEL_TC_CV(1), &tmp); + } while (upper != tmp); + + return (upper << 16) | lower; +} Is: +static u64 tc_get_cycles(struct clocksource *cs) +{ + u32 lower, upper, tmp; + + do { + regmap_read(tc.regmap, ATMEL_TC_CV(1), &upper); lock(); lot-of-things(); unlock(); + regmap_read(tc.regmap, ATMEL_TC_CV(0), &lower); lock(); lot-of-things(); unlock(); + regmap_read(tc.regmap, ATMEL_TC_CV(1), &tmp); lock(); lot-of-things(); unlock(); + } while (upper != tmp); + + return (upper << 16) | lower; +} I suggest to look what is in 'lot-of-things()' and especially what is doing regcache_read(). May be you can reconsider the regmap? This driver is the only one use the regmap AFAICT and I don't think it is adequate. > > > > Can you explain why we have two clocks here? > > > > > > > > > > Each channel have its clock, I can add a comment if you want. > > > > I don't understand. Why do we have two clocks? > > > > One channel is driven by one clock and the second one takes the overflow signal > > from the first one, so no second clock is involved there, no? > > > > Those are the peripheral clocks, they are not used by the counters but > used to be able to read/write the registers. Mmh, strange. Why is the clk[0]'s rate used in this case? -- Linaro.org ? Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog