On 12/16/13 21:17, Tony Lindgren wrote: > * Igor Grinberg <grinberg@xxxxxxxxxxxxxx> [131216 05:57]: >> On 12/13/13 21:22, Tony Lindgren wrote: >> >> SB-T35 has only one SMSC Ethernet on-board (smsc2), >> while the first one is on the cm-t3530 and cm-t3730 modules. >> SBC-T3517 has only one _SMSC_ Ethernet and it is on the SB-T35 base board. >> cm-t3517 has EMAC Ethernet on-board instead of the SMSC. > > OK, I'll move that. > >> There is also another base board - CB-T3, which turns into >> EM-T3530 and EM-T3730 once you plug the cm-t3530 or cm-t3730 into CB-T3. >> >> http://compulab.co.il/products/handheld-computers/em-t3730/ >> http://compulab.co.il/products/handheld-computers/em-t3530/ >> >> The CB-T3 base board does not have any Ethernet controllers on it, >> so the EM-T3x systems have only one Ethernet controller - the one on the >> CPU board - SMSC. > > OK > >> This makes it four boards: >> sb-t35 - base board has only one Ethernet controller (SMSC) >> cb-t3 - base board has no Ethernet controllers >> cm-t3530 - CPU board (plugged into sb-t35) has only one Ethernet controller (SMSC) >> cm-t3730 - CPU board (plugged into sb-t35) has only one Ethernet controller (SMSC) >> cm-t3537 - CPU board (plugged into sb-t35) has only one Ethernet controller (EMAC) >> >> The SBC-Txxx and EM-Txxx- means both boards connected (base board with CPU board). > > OK > >>> + >>> + cpus { >>> + cpu@0 { >>> + cpu0-supply = <&vcc>; >>> + }; >>> + }; >>> + >>> + leds { >>> + compatible = "gpio-leds"; >>> + ledb { >>> + label = "cm-t35:green"; >>> + gpios = <&gpio6 26 GPIO_ACTIVE_HIGH>; /* gpio186 */ >>> + linux,default-trigger = "heartbeat"; >> >> Although all CPU boards have the same GPIO routed to the heartbeat LED, >> this property is related to the CPU board and not to the base (mother) board. >> The same GPIO is used intensionally for s/w simplicity. >> If we are about to save code duplication, I'd like to have a common >> to CPU boards file, say omap3-cm-t3x.dtsi - that way it will better reflect >> the actual hardware. > > OK will move. > >>> +&i2c1 { >>> + clock-frequency = <2600000>; >>> + >>> + twl: twl@48 { >>> + reg = <0x48>; >>> + interrupts = <7>; /* SYS_NIRQ cascaded to intc */ >>> + interrupt-parent = <&intc>; >>> + }; >>> +}; >> >> This one should be a part of the common file. > > OK > >>> +#include "twl4030.dtsi" >>> +#include "twl4030_omap3.dtsi" >>> + >>> +&i2c3 { >>> + clock-frequency = <400000>; >>> +}; >>> + >>> +&mmc1 { >>> + vmmc-supply = <&vmmc1>; >>> + vmmc_aux-supply = <&vsim>; >> >> Is it essential to always provide vmmc_aux-supply property? >> In fact, the TPS65930 does not have the VSIM supply... >> It is a mistake in the upstream board file. > > OK > >> I fixed this locally a while ago, but haven't submitted upstream... >> So we should either leave this property unpopulated (if we can...) or >> provide a fixed regulator to populate it. >> >>> + bus-width = <8>; >> >> I'm not sure what this property should reflect. >> Both cm-t3530 and cm-t3730 use 4 bit MMC1 bus. > > Hmm except with the muxing it seems to work just fine with 8-bit :) Yes, this is because they are just not connected on sb-t35. > I noticed that by accident as I copied the entry from omap3-evm. > > So it seems that all the 8 lines are available for the MMC1, but can > be optionally also routed somewhere else? The concept of having CPU board and base (mother) board implies that our customers build a base board of their own. This way, the customer decides what to do with those lines. cm-t3x just makes them available on the connector, while neither sb-t35, nor cb-t3 uses them. > >>> +++ b/arch/arm/boot/dts/omap3-sbc-t3517.dts >>> @@ -0,0 +1,18 @@ >>> +/* >>> + * Suppport for CompuLab SBC-T3517 with CM-T3517 >>> + */ >>> +/dts-v1/; >>> + >>> +#include "am3517.dtsi" >>> +#include "omap3-sb-t35.dtsi" >>> + >>> + >>> +/ { >>> + model = "CompuLab SBC-T3517 with CM-T3517"; >>> + compatible = "compulab,omap3-sbc-t3517", "ti,am3517", "ti,omap3"; >>> + >>> + memory { >>> + device_type = "memory"; >>> + reg = <0x80000000 0x10000000>; /* 256 MB */ >> >> Can we support multiple memory sizes through static DT, or >> the only way is to update this property in the bootloader? > > Well we should probably have a minimal .dts file for the variants > too that can include omap3-cm-t37.dts and so on. I guess some > bootloaders are capable of rewriting the .dtb too. Ok. So what do you think, is it fine to have something like: omap3-cm-t3x.dtsi - common to cm-t3x cpu boards omap3-cm-t3x30.dtsi - common to cm-t3730 and cm-t3530 omap3-cm-t3730.dtsi - cm-t3730 specific omap3-cm-t3530.dtsi - cm-t3530 specific omap3-cm-t3517.dtsi - cm-t3517 specific omap3-sb-t35.dtsi - sb-t35 specific omap3-cb-t3.dtsi - cb-t3 specific omap3-sbc-t3730-256mb.dts - sb-t35 with cm-t3730 and 256MB memory size omap3-sbc-t3730-128mb.dts - sb-t35 with cm-t3730 and 128MB memory size omap3-sbc-t3730-64mb.dts - sb-t35 with cm-t3730 and 64MB memory size omap3-sbc-t3530-256mb.dts - sb-t35 with cm-t3530 and 256MB memory size omap3-sbc-t3530-128mb.dts - sb-t35 with cm-t3530 and 128MB memory size omap3-sbc-t3530-64mb.dts - sb-t35 with cm-t3530 and 64MB memory size omap3-sbc-t3517-256mb.dts - sb-t35 with cm-t3517 and 256MB memory size omap3-sbc-t3517-128mb.dts - sb-t35 with cm-t3517 and 128MB memory size omap3-em-t3730-256mb.dts - cb-t3 with cm-t3730 and 256MB memory size omap3-em-t3730-128mb.dts - cb-t3 with cm-t3730 and 128MB memory size omap3-em-t3730-64mb.dts - cb-t3 with cm-t3730 and 64MB memory size omap3-em-t3530-256mb.dts - cb-t3 with cm-t3530 and 256MB memory size omap3-em-t3530-128mb.dts - cb-t3 with cm-t3530 and 128MB memory size omap3-em-t3530-64mb.dts - cb-t3 with cm-t3530 and 64MB memory size or is it too much... ;-))) The above gives the full coverage of the 2x3 boards. I think we can drop the different memory sizes and let boot loader adjust the blob. This will make the list shorter: omap3-cm-t3x.dtsi - common to cm-t3x cpu boards omap3-cm-t3x30.dtsi - common to cm-t3730 and cm-t3530 omap3-cm-t3730.dtsi - cm-t3730 specific omap3-cm-t3530.dtsi - cm-t3530 specific omap3-cm-t3517.dtsi - cm-t3517 specific omap3-sb-t35.dtsi - sb-t35 specific omap3-cb-t3.dtsi - cb-t3 specific omap3-sbc-t3730.dts - sb-t35 with cm-t3730 and default memory size omap3-sbc-t3530.dts - sb-t35 with cm-t3530 and default memory size omap3-sbc-t3517.dts - sb-t35 with cm-t3517 and default memory size omap3-em-t3730.dts - cb-t3 with cm-t3730 and default memory size omap3-em-t3530.dts - cb-t3 with cm-t3530 and default memory size So what do you think? > >>> + mmc1_pins: pinmux_mmc1_pins { >>> + pinctrl-single,pins = < >>> + 0x114 (PIN_OUTPUT_PULLUP | MUX_MODE0) /* sdmmc1_clk.sdmmc1_clk */ >>> + 0x116 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_cmd.sdmmc1_cmd */ >>> + 0x118 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat0.sdmmc1_dat0 */ >>> + 0x11a (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat1.sdmmc1_dat1 */ >>> + 0x11c (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat2.sdmmc1_dat2 */ >>> + 0x11e (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat3.sdmmc1_dat3 */ >>> + 0x120 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat4.sdmmc1_dat4 */ >>> + 0x122 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat5.sdmmc1_dat5 */ >>> + 0x124 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat6.sdmmc1_dat6 */ >>> + 0x126 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc1_dat7.sdmmc1_dat7 */ >> >> The dat{4,5,6,7} pins are not used either on cm-t3530, or cm-t3730. > > But it seems to work and makes MMC1 faster :) Might be worth checking > though, maybe those pins have multiple optional routings available? They are routed to the connector of the base board and then... just not connected... Interesting, how that can make the MMC1 faster? Are you sure? > >>> + mmc2_pins: pinmux_mmc2_pins { >>> + pinctrl-single,pins = < >>> + 0x128 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc2_clk.sdmmc2_clk */ >>> + 0x12a (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc2_cmd.sdmmc2_cmd */ >>> + 0x12c (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc2_dat0.sdmmc2_dat0 */ >>> + 0x12e (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc2_dat1.sdmmc2_dat1 */ >>> + 0x130 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc2_dat2.sdmmc2_dat2 */ >>> + 0x132 (PIN_INPUT_PULLUP | MUX_MODE0) /* sdmmc2_dat3.sdmmc2_dat3 */ >> >> Here the following is missing: >> 0x134 (PIN_OUTPUT | MUX_MODE1) /* sdmmc2_dat4.sdmmc2_dir_dat0 */ > > That seems to be used for the wl12xx GPIO, so it's listed at > wl12xx_gpio below. Yes, but only on cm-t3730 (and actually starting from revision 1.2), not cm-t3530... > >>> + 0x136 (PIN_OUTPUT | MUX_MODE1) /* sdmmc2_dat5.sdmmc2_dir_dat1 */ >>> + 0x138 (PIN_OUTPUT | MUX_MODE1) /* sdmmc2_dat6.sdmmc2_dir_cmd */ >>> + 0x13a (PIN_INPUT | MUX_MODE1) /* sdmmc2_dat7.sdmmc2_clkin */ >> >> All four above pins (dat{4,5,6,7}) should be muxed only on cm-t3530. >> cm-t3730 uses the sdmmc2_dat4 (gpio136) for wl12xx irq, >> and does nor use the rest (dat{5,6,7}) at all. > > Hmm OK, maybe those extra pins also have alternative routings available? The same as with MMC1, they are routed to the connector and not routed at all on sb-t35. > >>> + wl12xx_gpio: pinmux_wl12xx_gpio { >>> + pinctrl-single,pins = < >>> + 0xb2 (PIN_OUTPUT | MUX_MODE4) /* dss_data3.gpio_73 */ >>> + 0x134 (PIN_INPUT | MUX_MODE4) /* sdmmc2_dat4.gpio_136 */ >>> + >; >>> + }; >>> +}; > > Regards, > > Tony > -- Regards, Igor. -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html