Hi Alban, thanks for aggregating all the background information about the issue. On Tue Jun 11, 2024 at 9:38 PM CEST, Alban Browaeys wrote: > > Could you test the enable-strobe property on all the dts that disabled > hs400es due to this new default (I lack the hardware to test the > patches). > At least arch/arm64/boot/dts/rockchip/rk3399-khadas-edge.dtsi > arch/arm64/boot/dts/rockchip/rk3399-rock-4c-plus.dts > arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi > and ping the board tester for other boards to test if they require > "rockchip,enable-strobe-pulldown" for EMMC HS400 write support (read is > fine even with the new default). > I tested some of the boards that include rk3399-rock-pi-4.dtsi (see below). > On Tue, 2024-02-27 at 10:11 +0000, Folker Schwesinger wrote: > > with the following applied, the EMMC related errors are gone. dmesg > > only > > shows "Purging ... bytes" during my tests: > > > > diff --git a/arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi > > b/arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi > > index f2279aa6ca9e..ae0fb87e1a8b 100644 > > --- a/arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi > > +++ b/arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi > > @@ -647,8 +647,10 @@ &saradc { > > &sdhci { > > max-frequency = <150000000>; > > bus-width = <8>; > > - mmc-hs200-1_8v; > > + mmc-hs400-1_8v; > > + mmc-hs400-enhanced-strobe; > > non-removable; > > + rockchip,enable-strobe-pulldown; > > status = "okay"; > > }; > > > > For testing I ran dd three times in a row: > > > > dd if=/dev/zero of=./zero.bin bs=1M count=5000 > > > > I tested this on both a Rock 4SE board and a Rock Pi 4B+ board with > > the > > same results. > > Folker, are you confident "Rock 4SE board and Rock Pi 4B+" were fixed > with above patch? > Ie the "rockchip,enable-strobe-pulldown;" should be under an > "rockchip,rk3399-emmc-phy" compaible node, that is &emmc_phy in > arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi, not sdhci. > The above diff was just a quick shot at testing the "rockchip,enable-strobe-pulldown" property when I first learned about it. I later realized that the property belongs under the &emmc_phy node as you suggested. That's what I did in the other patchset I sent a bit later, which was accepted and applied: https://lists.infradead.org/pipermail/linux-rockchip/2024-March/045723.html f720dd9b8b6d8b2160beda789429d5489ce8a099 c1b1f340dd7db11f273e426e110697551c9f501f So yes, the Rock 4SE, Rock Pi 4B and Rock Pi 4B+ boards all were fixed with the patch. I regularly have the Rock 4SE and Rock 4B running from EMMC, always with the patch applied since I sent it.