From mboxrd@z Thu Jan 1 00:00:00 1970 From: Geert Uytterhoeven Subject: Re: [PATCH v5 00/26] Fix watchdog on Renesas R-Car Gen2 and RZ/G1 Date: Tue, 20 Feb 2018 13:51:28 +0100 Message-ID: References: <1518457475-4480-1-git-send-email-fabrizio.castro@bp.renesas.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1518457475-4480-1-git-send-email-fabrizio.castro@bp.renesas.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=m.gmane.org@lists.infradead.org To: Fabrizio Castro Cc: Mark Rutland , Geert Uytterhoeven , Catalin Marinas , Michael Turquette , Will Deacon , Wolfram Sang , Wim Van Sebroeck , linux-clk , Magnus Damm , Russell King , Guenter Roeck , "open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" , Chris Paterson , Linux Watchdog Mailing List , Biju Das , Simon Horman , linux-arm-kernel@lists.infradead.org, Stephen Boyd , Linux-Renesas , Rob Herring , Philipp List-Id: devicetree@vger.kernel.org On Mon, Feb 12, 2018 at 6:44 PM, Fabrizio Castro wrote: > this series has been around for some time as RFC, and it has collected > useful comments from the community along the way. > The solution proposed by this patch set works for most R-Car Gen2 and > RZ/G1 devices, but not all of them. We now know that for some R-Car > Gen2 early revisions there is no proper software fix. Anyway, no > product has been built around early revisions, but development boards > mounting early revisions (basically prototypes) are still out there. > As a result, this series isn't enabling the internal watchdog on R-Car > Gen2 boards, developers may enable it in board specific device trees > if needed. > This series has been tested by me on the iwg20d, iwg22d, Lager, Alt, > and Koelsch boards. > > The problem > =========== > To deal with SMP on R-Car Gen2 and RZ/G1, we install a reset vector > to ICRAM1 and we program the [S]BAR registers so that when we turn ON > the non-boot CPUs they are redirected to the reset vector installed by > Linux in ICRAM1, and eventually they continue the execution to RAM, > where the SMP bring-up code will take care of the rest. > The content of the [S]BAR registers survives a watchdog triggered reset, > and as such after the watchdog fires the boot core will try and execute > the SMP bring-up code instead of jumping to the bootrom code. > > The fix > ======= > The main strategy for the solution is to let the reset vector decide > if it needs to jump to shmobile_boot_fn or to the bootrom code. > In a watchdog triggered reset scenario, since the [S]BAR registers keep > their values, the boot CPU will jump into the newly designed reset > vector, the assembly routine will eventually test WOVF (a bit in register > RWTCSRA that indicates if the watchdog counter has overflown, the value > of this bit gets retained in this scenario), and jump to the bootrom code > which will in turn load up the bootloader, etc. > When bringing up SMP or using CPU hotplug, the reset vector will jump > to shmobile_boot_fn instead. > > Thank you All for your help. > > Best regards, > > Fabrizio Castro (26): > ARM: shmobile: Add watchdog support > ARM: dts: r8a7743: Adjust SMP routine size > ARM: dts: r8a7745: Adjust SMP routine size > ARM: dts: r8a7790: Adjust SMP routine size > ARM: dts: r8a7791: Adjust SMP routine size > ARM: dts: r8a7792: Adjust SMP routine size > ARM: dts: r8a7793: Adjust SMP routine size > ARM: dts: r8a7794: Adjust SMP routine size > soc: renesas: rcar-rst: Enable watchdog as reset trigger for Gen2 > ARM: shmobile: rcar-gen2: Add watchdog support > dt-bindings: watchdog: renesas-wdt: Add R-Car Gen2 support > watchdog: renesas_wdt: Add R-Car Gen2 support > watchdog: renesas_wdt: Add restart handler > ARM: shmobile: defconfig: Enable RENESAS_WDT_GEN > clk: renesas: r8a7743: Add rwdt clock > clk: renesas: r8a7745: Add rwdt clock > clk: renesas: r8a7790: Add rwdt clock > clk: renesas: r8a7791/r8a7793: Add rwdt clock > clk: renesas: r8a7794: Add rwdt clock > ARM: dts: r8a7743: Add watchdog support to SoC dtsi > ARM: dts: r8a7745: Add watchdog support to SoC dtsi > ARM: dts: r8a7790: Add watchdog support to SoC dtsi > ARM: dts: r8a7791: Add watchdog support to SoC dtsi > ARM: dts: r8a7794: Add watchdog support to SoC dtsi > ARM: dts: iwg20m: Add watchdog support to SoM dtsi > ARM: dts: iwg22m: Add watchdog support to SoM dtsi > > .../devicetree/bindings/watchdog/renesas-wdt.txt | 19 ++++++-- > arch/arm/boot/dts/r8a7743-iwg20m.dtsi | 5 ++ > arch/arm/boot/dts/r8a7743.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7745-iwg22m.dtsi | 5 ++ > arch/arm/boot/dts/r8a7745.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7790.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7791.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7792.dtsi | 2 +- > arch/arm/boot/dts/r8a7793.dtsi | 2 +- > arch/arm/boot/dts/r8a7794.dtsi | 12 ++++- > arch/arm/configs/shmobile_defconfig | 1 + > arch/arm/mach-shmobile/common.h | 6 +++ > arch/arm/mach-shmobile/headsmp.S | 55 ++++++++++++++++++++++ > arch/arm/mach-shmobile/platsmp-apmu.c | 1 + > arch/arm/mach-shmobile/pm-rcar-gen2.c | 15 ++++-- > drivers/clk/renesas/r8a7743-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7745-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7790-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7791-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7794-cpg-mssr.c | 2 + > drivers/soc/renesas/rcar-rst.c | 35 +++++++++++--- > drivers/watchdog/renesas_wdt.c | 39 +++++++++++++-- Thanks, I've queued the clock patches in clk-renesas-for-v4.17, as they are a hard dependency for: (1) The new reset vector in arch/arm/mach-shmobile/ (2) The watchdog driver. Note that the watchdog driver itself (2) has a hard dependency on the new reset vector (1). Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qk0-f169.google.com ([209.85.220.169]:46225 "EHLO mail-qk0-f169.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751551AbeBTMva (ORCPT ); Tue, 20 Feb 2018 07:51:30 -0500 MIME-Version: 1.0 In-Reply-To: <1518457475-4480-1-git-send-email-fabrizio.castro@bp.renesas.com> References: <1518457475-4480-1-git-send-email-fabrizio.castro@bp.renesas.com> From: Geert Uytterhoeven Date: Tue, 20 Feb 2018 13:51:28 +0100 Message-ID: Subject: Re: [PATCH v5 00/26] Fix watchdog on Renesas R-Car Gen2 and RZ/G1 To: Fabrizio Castro Cc: Philipp Zabel , Rob Herring , Mark Rutland , Wim Van Sebroeck , Russell King , Catalin Marinas , Will Deacon , Michael Turquette , Stephen Boyd , Simon Horman , Magnus Damm , Geert Uytterhoeven , Wolfram Sang , Guenter Roeck , "open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" , Linux Watchdog Mailing List , Linux-Renesas , linux-arm-kernel@lists.infradead.org, linux-clk , Chris Paterson , Biju Das , Ramesh Shanmugasundaram Content-Type: text/plain; charset="UTF-8" Sender: linux-watchdog-owner@vger.kernel.org List-Id: linux-watchdog@vger.kernel.org On Mon, Feb 12, 2018 at 6:44 PM, Fabrizio Castro wrote: > this series has been around for some time as RFC, and it has collected > useful comments from the community along the way. > The solution proposed by this patch set works for most R-Car Gen2 and > RZ/G1 devices, but not all of them. We now know that for some R-Car > Gen2 early revisions there is no proper software fix. Anyway, no > product has been built around early revisions, but development boards > mounting early revisions (basically prototypes) are still out there. > As a result, this series isn't enabling the internal watchdog on R-Car > Gen2 boards, developers may enable it in board specific device trees > if needed. > This series has been tested by me on the iwg20d, iwg22d, Lager, Alt, > and Koelsch boards. > > The problem > =========== > To deal with SMP on R-Car Gen2 and RZ/G1, we install a reset vector > to ICRAM1 and we program the [S]BAR registers so that when we turn ON > the non-boot CPUs they are redirected to the reset vector installed by > Linux in ICRAM1, and eventually they continue the execution to RAM, > where the SMP bring-up code will take care of the rest. > The content of the [S]BAR registers survives a watchdog triggered reset, > and as such after the watchdog fires the boot core will try and execute > the SMP bring-up code instead of jumping to the bootrom code. > > The fix > ======= > The main strategy for the solution is to let the reset vector decide > if it needs to jump to shmobile_boot_fn or to the bootrom code. > In a watchdog triggered reset scenario, since the [S]BAR registers keep > their values, the boot CPU will jump into the newly designed reset > vector, the assembly routine will eventually test WOVF (a bit in register > RWTCSRA that indicates if the watchdog counter has overflown, the value > of this bit gets retained in this scenario), and jump to the bootrom code > which will in turn load up the bootloader, etc. > When bringing up SMP or using CPU hotplug, the reset vector will jump > to shmobile_boot_fn instead. > > Thank you All for your help. > > Best regards, > > Fabrizio Castro (26): > ARM: shmobile: Add watchdog support > ARM: dts: r8a7743: Adjust SMP routine size > ARM: dts: r8a7745: Adjust SMP routine size > ARM: dts: r8a7790: Adjust SMP routine size > ARM: dts: r8a7791: Adjust SMP routine size > ARM: dts: r8a7792: Adjust SMP routine size > ARM: dts: r8a7793: Adjust SMP routine size > ARM: dts: r8a7794: Adjust SMP routine size > soc: renesas: rcar-rst: Enable watchdog as reset trigger for Gen2 > ARM: shmobile: rcar-gen2: Add watchdog support > dt-bindings: watchdog: renesas-wdt: Add R-Car Gen2 support > watchdog: renesas_wdt: Add R-Car Gen2 support > watchdog: renesas_wdt: Add restart handler > ARM: shmobile: defconfig: Enable RENESAS_WDT_GEN > clk: renesas: r8a7743: Add rwdt clock > clk: renesas: r8a7745: Add rwdt clock > clk: renesas: r8a7790: Add rwdt clock > clk: renesas: r8a7791/r8a7793: Add rwdt clock > clk: renesas: r8a7794: Add rwdt clock > ARM: dts: r8a7743: Add watchdog support to SoC dtsi > ARM: dts: r8a7745: Add watchdog support to SoC dtsi > ARM: dts: r8a7790: Add watchdog support to SoC dtsi > ARM: dts: r8a7791: Add watchdog support to SoC dtsi > ARM: dts: r8a7794: Add watchdog support to SoC dtsi > ARM: dts: iwg20m: Add watchdog support to SoM dtsi > ARM: dts: iwg22m: Add watchdog support to SoM dtsi > > .../devicetree/bindings/watchdog/renesas-wdt.txt | 19 ++++++-- > arch/arm/boot/dts/r8a7743-iwg20m.dtsi | 5 ++ > arch/arm/boot/dts/r8a7743.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7745-iwg22m.dtsi | 5 ++ > arch/arm/boot/dts/r8a7745.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7790.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7791.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7792.dtsi | 2 +- > arch/arm/boot/dts/r8a7793.dtsi | 2 +- > arch/arm/boot/dts/r8a7794.dtsi | 12 ++++- > arch/arm/configs/shmobile_defconfig | 1 + > arch/arm/mach-shmobile/common.h | 6 +++ > arch/arm/mach-shmobile/headsmp.S | 55 ++++++++++++++++++++++ > arch/arm/mach-shmobile/platsmp-apmu.c | 1 + > arch/arm/mach-shmobile/pm-rcar-gen2.c | 15 ++++-- > drivers/clk/renesas/r8a7743-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7745-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7790-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7791-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7794-cpg-mssr.c | 2 + > drivers/soc/renesas/rcar-rst.c | 35 +++++++++++--- > drivers/watchdog/renesas_wdt.c | 39 +++++++++++++-- Thanks, I've queued the clock patches in clk-renesas-for-v4.17, as they are a hard dependency for: (1) The new reset vector in arch/arm/mach-shmobile/ (2) The watchdog driver. Note that the watchdog driver itself (2) has a hard dependency on the new reset vector (1). Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds From mboxrd@z Thu Jan 1 00:00:00 1970 From: geert@linux-m68k.org (Geert Uytterhoeven) Date: Tue, 20 Feb 2018 13:51:28 +0100 Subject: [PATCH v5 00/26] Fix watchdog on Renesas R-Car Gen2 and RZ/G1 In-Reply-To: <1518457475-4480-1-git-send-email-fabrizio.castro@bp.renesas.com> References: <1518457475-4480-1-git-send-email-fabrizio.castro@bp.renesas.com> Message-ID: To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Mon, Feb 12, 2018 at 6:44 PM, Fabrizio Castro wrote: > this series has been around for some time as RFC, and it has collected > useful comments from the community along the way. > The solution proposed by this patch set works for most R-Car Gen2 and > RZ/G1 devices, but not all of them. We now know that for some R-Car > Gen2 early revisions there is no proper software fix. Anyway, no > product has been built around early revisions, but development boards > mounting early revisions (basically prototypes) are still out there. > As a result, this series isn't enabling the internal watchdog on R-Car > Gen2 boards, developers may enable it in board specific device trees > if needed. > This series has been tested by me on the iwg20d, iwg22d, Lager, Alt, > and Koelsch boards. > > The problem > =========== > To deal with SMP on R-Car Gen2 and RZ/G1, we install a reset vector > to ICRAM1 and we program the [S]BAR registers so that when we turn ON > the non-boot CPUs they are redirected to the reset vector installed by > Linux in ICRAM1, and eventually they continue the execution to RAM, > where the SMP bring-up code will take care of the rest. > The content of the [S]BAR registers survives a watchdog triggered reset, > and as such after the watchdog fires the boot core will try and execute > the SMP bring-up code instead of jumping to the bootrom code. > > The fix > ======= > The main strategy for the solution is to let the reset vector decide > if it needs to jump to shmobile_boot_fn or to the bootrom code. > In a watchdog triggered reset scenario, since the [S]BAR registers keep > their values, the boot CPU will jump into the newly designed reset > vector, the assembly routine will eventually test WOVF (a bit in register > RWTCSRA that indicates if the watchdog counter has overflown, the value > of this bit gets retained in this scenario), and jump to the bootrom code > which will in turn load up the bootloader, etc. > When bringing up SMP or using CPU hotplug, the reset vector will jump > to shmobile_boot_fn instead. > > Thank you All for your help. > > Best regards, > > Fabrizio Castro (26): > ARM: shmobile: Add watchdog support > ARM: dts: r8a7743: Adjust SMP routine size > ARM: dts: r8a7745: Adjust SMP routine size > ARM: dts: r8a7790: Adjust SMP routine size > ARM: dts: r8a7791: Adjust SMP routine size > ARM: dts: r8a7792: Adjust SMP routine size > ARM: dts: r8a7793: Adjust SMP routine size > ARM: dts: r8a7794: Adjust SMP routine size > soc: renesas: rcar-rst: Enable watchdog as reset trigger for Gen2 > ARM: shmobile: rcar-gen2: Add watchdog support > dt-bindings: watchdog: renesas-wdt: Add R-Car Gen2 support > watchdog: renesas_wdt: Add R-Car Gen2 support > watchdog: renesas_wdt: Add restart handler > ARM: shmobile: defconfig: Enable RENESAS_WDT_GEN > clk: renesas: r8a7743: Add rwdt clock > clk: renesas: r8a7745: Add rwdt clock > clk: renesas: r8a7790: Add rwdt clock > clk: renesas: r8a7791/r8a7793: Add rwdt clock > clk: renesas: r8a7794: Add rwdt clock > ARM: dts: r8a7743: Add watchdog support to SoC dtsi > ARM: dts: r8a7745: Add watchdog support to SoC dtsi > ARM: dts: r8a7790: Add watchdog support to SoC dtsi > ARM: dts: r8a7791: Add watchdog support to SoC dtsi > ARM: dts: r8a7794: Add watchdog support to SoC dtsi > ARM: dts: iwg20m: Add watchdog support to SoM dtsi > ARM: dts: iwg22m: Add watchdog support to SoM dtsi > > .../devicetree/bindings/watchdog/renesas-wdt.txt | 19 ++++++-- > arch/arm/boot/dts/r8a7743-iwg20m.dtsi | 5 ++ > arch/arm/boot/dts/r8a7743.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7745-iwg22m.dtsi | 5 ++ > arch/arm/boot/dts/r8a7745.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7790.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7791.dtsi | 12 ++++- > arch/arm/boot/dts/r8a7792.dtsi | 2 +- > arch/arm/boot/dts/r8a7793.dtsi | 2 +- > arch/arm/boot/dts/r8a7794.dtsi | 12 ++++- > arch/arm/configs/shmobile_defconfig | 1 + > arch/arm/mach-shmobile/common.h | 6 +++ > arch/arm/mach-shmobile/headsmp.S | 55 ++++++++++++++++++++++ > arch/arm/mach-shmobile/platsmp-apmu.c | 1 + > arch/arm/mach-shmobile/pm-rcar-gen2.c | 15 ++++-- > drivers/clk/renesas/r8a7743-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7745-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7790-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7791-cpg-mssr.c | 2 + > drivers/clk/renesas/r8a7794-cpg-mssr.c | 2 + > drivers/soc/renesas/rcar-rst.c | 35 +++++++++++--- > drivers/watchdog/renesas_wdt.c | 39 +++++++++++++-- Thanks, I've queued the clock patches in clk-renesas-for-v4.17, as they are a hard dependency for: (1) The new reset vector in arch/arm/mach-shmobile/ (2) The watchdog driver. Note that the watchdog driver itself (2) has a hard dependency on the new reset vector (1). Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds