On 2015-05-22 13:11, Arnd Bergmann wrote: > On Friday 22 May 2015 16:21:54 Sanchayan Maity wrote: >> +#define OCOTP_CFG0_OFFSET 0x00000410 >> +#define OCOTP_CFG1_OFFSET 0x00000420 >> +#define MSCM_CPxCOUNT_OFFSET 0x0000002C >> +#define MSCM_CPxCFG1_OFFSET 0x00000014 >> +#define ROM_REVISION_OFFSET 0x00000080 >> + >> +static const struct of_device_id vf610_soc_bus_match[] = { >> + { .compatible = "fsl,vf610-mscm-cpucfg", }, >> + { /* sentinel */ } >> +}; >> + >> +static int __init vf610_soc_init(void) >> +{ >> + struct regmap *ocotp_regmap, *mscm_regmap, *rom_regmap; >> + struct soc_device_attribute *soc_dev_attr; >> + struct soc_device *soc_dev; >> + struct device_node *np; >> + char soc_type[] = "xx0"; >> + u32 cpxcount, cpxcfg1; >> + u32 soc_id1, soc_id2, rom_rev; >> + u64 soc_id; >> + int ret; >> + >> + np = of_find_matching_node(NULL, vf610_soc_bus_match); >> + if (!np) >> + return -ENODEV; >> + > > Why not use module_platform_driver() and make this a probe function instead? > >> + ocotp_regmap = syscon_regmap_lookup_by_compatible("fsl,vf610-ocotp"); >> + if (IS_ERR(ocotp_regmap)) { >> + pr_err("regmap lookup for octop failed\n"); >> + return PTR_ERR(ocotp_regmap); >> + } >> + >> + mscm_regmap = syscon_regmap_lookup_by_compatible("fsl,vf610-mscm-cpucfg"); >> + if (IS_ERR(mscm_regmap)) { >> + pr_err("regmap lookup for mscm failed"); >> + return PTR_ERR(mscm_regmap); >> + } >> + >> + rom_regmap = syscon_regmap_lookup_by_compatible("fsl,vf610-ocrom"); >> + if (IS_ERR(rom_regmap)) { >> + pr_err("regmap lookup for ocrom failed"); >> + return PTR_ERR(rom_regmap); >> + } > > Can you use syscon_regmap_lookup_by_phandle instead, and put the > phandles in the fsl,vf610-mscm-cpucfg node? Hm, with that we would wire up hardware modules which does nothing has to do with each other. We just happen to need a driver which collects information accross the SoC. I'm not sure we should put the modules required into the device tree. I don't think its nice to have the compatible strings in the source code, however it feels more appropriate than in the device tree, IMHO... > Also, I'd argue that the mscm should not be a syscon device at all, > but instead I'd use platform_get_resource()/devm_ioremap_resource() > to get an __iomem pointer. We need to have mscm-cpucfg to be syscon because we need to get the CPU personality in the MSCM interrupt router driver (irq-vf610-mscm-ir.c). -- Stefan -- 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