On Thu, Feb 08, 2018 at 11:34:06AM +0100, Geert Uytterhoeven wrote: > Hi all, > > This patch series enables the builtin watchdog timer on R-Car Gen2 SoCs > on all supported boards, and builds on top of Fabrizio's "[RFC v4 00/26] > Fix watchdog on Renesas R-Car Gen2 and RZ/G1". It is marked RFC as it > is known not to work on all SoCs and SoC revisions. > > Based on my experiments, there are 3 success/failure modes: > > 1. It works! > > This is the case on R-Car M2-N ES1.0 and E2 ES1.0 (tested on gose > and alt). > > 2. The system hangs on watchdog timeout, unless only the first CPU > core is in use. > This can be achieved by booting with "maxcpus=1", or by manually > offlining all secondary CPU cores first: > > for i in /sys/*/*/cpu/cpu[1-9]*; do echo 0 > $i/online; done > > This is the case on R-Car H2 ES1.0 and M2-W ES1.0 (tested on lager > and koelsch). > > 3. The system hangs on watchdog timeout, unless the kernel was built > with CONFIG_SMP disabled. > > This is the case on R-Car V2H ES1.1 (tested on blanche). > Presumably the restart part in the new shmobile_boot_vector_gen2 is > not compatible with R-Car V2H in some way (CPU off/onlining does > work). > > Once we know on which SoCs and revisions it does work, we can add the > proper checks. Right now I'm thinking of adding something like: > > if (IS_ENABLED(CONFIG_SMP) && maxcpus > 1 && > soc_device_match(rcar_gen2_wdt_needs_up)) > return -ENODEV; > > to the watchdog's probe function (assumed we can fix > shmobile_boot_vector_gen2 for R-Car V2H, else we need more checks). > > For your convenience, Fabrizio's and my series are available in the > topic/rcar-gen2-wdt-v4 resp. topic/rcar-gen2-wdt-v4+ branches of my > renesas-drivers git repository at > git://git.kernel.org/pub/scm/linux/kernel/git/geert/renesas-drivers.git. > > shmobile_defconfig should enable all support needed for testing. > > Test procedure: > > cat > /dev/watchdog0 & for i in $(seq 100); do echo $i; sleep 1; done > > If everything goes well, the board should reboot after 60 seconds. > > Note that on several boards there are multiple watchdogs (RWDT and > DA9063 PMIC): > > # ls -l /dev/watchdog* > crw------- 1 root root 10, 130 Jan 1 2000 /dev/watchdog > crw------- 1 root root 247, 0 Jan 1 2000 /dev/watchdog0 > crw------- 1 root root 247, 1 Jan 1 2000 /dev/watchdog1 > > /dev/watchdog always corresponds to /dev/watchdog0. > The first one is usually the RWDT. If you want to be 100% sure, look in > sysfs for devices with the (dynamic!) major matching the /dev/watchdog* > files: > > # find /sys/devices/ -name dev | xargs grep 247: > /sys/devices/platform/soc/e6020000.watchdog/watchdog/watchdog0/dev:247:0 > /sys/devices/platform/soc/e60b0000.i2c/i2c-6/6-0058/da9063-watchdog/watchdog/watchdog1/dev:247:1 > > Thanks for testing, especially on newer revisions of R-Car H2, M2-W, and > V2H, and on other boards! > > Geert Uytterhoeven (11): > clk: renesas: r8a7792: Add rwdt clock > ARM: dts: r8a7792: Add RWDT node > ARM: dts: r8a7793: Add RWDT node > ARM: dts: lager: Enable watchdog support > ARM: dts: koelsch: Enable watchdog support > ARM: dts: porter: Enable watchdog support > ARM: dts: blanche: Enable watchdog support > ARM: dts: wheat: Enable watchdog support > ARM: dts: gose: Enable watchdog support > ARM: dts: alt: Enable watchdog support > ARM: dts: silk: Enable watchdog support Thanks, "dts" patches applied for v4.17.
