On 20/12/2021 12:51, Roger Quadros wrote: > > > On 20/12/2021 13:05, Krzysztof Kozlowski wrote: >> On 20/12/2021 11:53, Roger Quadros wrote: >>> >>> >>> On 17/12/2021 17:21, Krzysztof Kozlowski wrote: >>>> On 17/12/2021 11:29, Roger Quadros wrote: >>>>> As more compatibles can be added to the GPMC NAND controller driver >>>>> use a compatible match table. >>>>> >>>>> Signed-off-by: Roger Quadros <rogerq@xxxxxxxxxx> >>>>> --- >>>>> drivers/memory/omap-gpmc.c | 8 +++++++- >>>>> drivers/mtd/nand/raw/omap2.c | 2 +- >>>>> include/linux/platform_data/mtd-nand-omap2.h | 5 +++++ >>>>> 3 files changed, 13 insertions(+), 2 deletions(-) >>>>> >>>>> diff --git a/drivers/memory/omap-gpmc.c b/drivers/memory/omap-gpmc.c >>>>> index 624153048182..814ddb45c13d 100644 >>>>> --- a/drivers/memory/omap-gpmc.c >>>>> +++ b/drivers/memory/omap-gpmc.c >>>>> @@ -2091,6 +2091,7 @@ static int gpmc_probe_generic_child(struct platform_device *pdev, >>>>> u32 val; >>>>> struct gpio_desc *waitpin_desc = NULL; >>>>> struct gpmc_device *gpmc = platform_get_drvdata(pdev); >>>>> + bool is_nand = false; >>>>> >>>>> if (of_property_read_u32(child, "reg", &cs) < 0) { >>>>> dev_err(&pdev->dev, "%pOF has no 'reg' property\n", >>>>> @@ -2183,7 +2184,12 @@ static int gpmc_probe_generic_child(struct platform_device *pdev, >>>>> } >>>>> } >>>>> >>>>> - if (of_device_is_compatible(child, "ti,omap2-nand")) { >>>>> +#if defined(CONFIG_MTD_NAND_OMAP2) >>>> >>>> if (IS_ENABLED()) is preferred. If needed, you could make omap_nand_ids >>>> symbol visible always (so without ifdef around it), because extern >>>> structure should not have impact when not defined (if I recall >>>> correctly...). >>>> >>>>> + if (of_match_node(omap_nand_ids, child)) >>>>> + is_nand = true; >>>>> +#endif >>>>> + >>>>> + if (is_nand) { >>>>> /* NAND specific setup */ >>>>> val = 8; >>>>> of_property_read_u32(child, "nand-bus-width", &val); >>>>> diff --git a/drivers/mtd/nand/raw/omap2.c b/drivers/mtd/nand/raw/omap2.c >>>>> index b26d4947af02..fff834ee726f 100644 >>>>> --- a/drivers/mtd/nand/raw/omap2.c >>>>> +++ b/drivers/mtd/nand/raw/omap2.c >>>>> @@ -2352,7 +2352,7 @@ static int omap_nand_remove(struct platform_device *pdev) >>>>> return ret; >>>>> } >>>>> >>>>> -static const struct of_device_id omap_nand_ids[] = { >>>>> +const struct of_device_id omap_nand_ids[] = { >>>>> { .compatible = "ti,omap2-nand", }, >>>>> {}, >>>>> }; >>>> >>>> I think OMAP2 NAND driver can be a module, so this should have >>>> EXPORT_SYMBOL. >>> >>> To make it work in all combinations (e.g. omap_gpmc built in and >>> nand/raw/omap2.c as module) I had to define omap_nand_ids table as static >>> in the linux/platform_data/mtd-nand-omap2.h header. >>> >>> EXPORT_SYMBOL will of course be not required there. ;) >>> >> Which case exactly does it require to be static in the header? > > Maybe there is a better way to do it. > e.g. If omap_gpmc.c is built-in and nand/raw/omap2.c is not built or built as > a module, what better way we can use? Ah, you are right, that is the tricky configuration. It could be a separate built-in source file which is being selected by OMAP_GPMC and MTD_NAND_OMAP2, but it would be also an overkill for one-item array. I guess your solution with header is the easiest. Best regards, Krzysztof