On Mon, Mar 18, 2019 at 07:38:26PM +0900, Chanwoo Choi wrote: > Thanks for comment. I add my comments > and then you have to rebase it on latest v5.0-rc1 > because the merge conflict happen on v5.0-rc1.
Thanks for review, see my answers below. Non-answered items will be fixed accordingly. > >> +config EXTCON_INTEL_MRFLD > > > >> + tristate "Intel MErrifield Basin Cove PMIC extcon driver" > > > > ME -> Me (will be fixed) > > > >> + depends on INTEL_SOC_PMIC_MRFLD > > This driver uses the regmap interface. So, you better to add > following dependency? > - select REGMAP_I2C or REGMAP_SPI None of them fits this or MFD driver. See below. > But, if 'INTEL_SOC_PMIC_MRFLE' selects already REGMAP_* > configuration. It is not necessary. https://lore.kernel.org/lkml/[email protected]/ It selects REGMAP_IRQ which selects necessary bits from regmap API. > >> + help > >> + Say Y here to enable extcon support for charger detection / control > >> + on the Intel Merrifiel Basin Cove PMIC. > > What is correct word? > - Merrifield? is used on above > - Merrifiel? Merrifield is a correct one. Thanks for spotting this. > >> +static int mrfld_extcon_set(struct mrfld_extcon_data *data, unsigned int > >> reg, > >> + unsigned int mask) > >> +{ > >> + return regmap_update_bits(data->regmap, reg, mask, 0xff); > >> +} > > mrfld_extcon_clear() and mrfld_extcon_set() are just wrapper function > for regmap interface. I think that you better to define > the meaningful defintion for '0x00' and '0xff' as following: > > (just example, you may make the more correct name) > #define INTEL_MRFLD_RESET 0x00 > #define INTEL_MRFLD_SET 0xff It makes a little sense here, the idea is to reduce parameters. I could ideally write (..., mask, ~mask) for clear and (..., mask, mask) for set > And then you better to use the 'regmap_update_bits()' function > directly instead of mrfld_extcon_clear/set'. It will bring duplication of long definitions and reduce readability of the code. > >> + /* > >> + * It seems SCU firmware clears the content of BCOVE_CHGRIRQ1 > >> + * and makes it useless for OS. Instead we compare a previously > >> + * stored status to the current one, provided by BCOVE_SCHGRIRQ1. > >> + */ > >> + ret = regmap_read(regmap, BCOVE_SCHGRIRQ1, &status); > >> + if (ret) > >> + return ret; > >> + > >> + if (!(status ^ data->status)) > >> + return -ENODATA; > >> + > >> + if ((status ^ data->status) & BCOVE_CHGRIRQ_USBIDDET) > >> + ret = mrfld_extcon_role_detect(data); > This line gets the return value from mrfld_extcon_role_detect(data) > without any error handling and then the below line just saves 'status' > to 'data->status' regardless of 'ret' value. > > I think that you have to handle the error case of > 'ret = mrfld_extcon_role_detect(data)'. I'm not sure of the consequences of such change. I will give it some tests, and then will proceed accordingly. > >> + .name = KBUILD_MODNAME, > > Where is the definition of KBUILD_MODNAME? Are you missing? In the Makefile. Nothing is missed here. But I could put its content explicitly here. -- With Best Regards, Andy Shevchenko
